#  @ryan_kidd44 Ryan Kidd Ryan Kidd posts on X about ai, open ai, applications, agi the most. They currently have [-----] followers and [---] posts still getting attention that total [-----] engagements in the last [--] hours. ### Engagements: [-----] [#](/creator/twitter::1102399276334759936/interactions)  - [--] Week [-----] -55% - [--] Month [------] +209% - [--] Months [-------] +35,472% - [--] Year [-------] +614% ### Mentions: [--] [#](/creator/twitter::1102399276334759936/posts_active)  ### Followers: [-----] [#](/creator/twitter::1102399276334759936/followers)  - [--] Week [-----] +0.31% - [--] Month [-----] +6.90% - [--] Months [-----] +52% - [--] Year [-----] +101% ### CreatorRank: [---------] [#](/creator/twitter::1102399276334759936/influencer_rank)  ### Social Influence **Social category influence** [technology brands](/list/technology-brands) [travel destinations](/list/travel-destinations) [stocks](/list/stocks) [finance](/list/finance) [countries](/list/countries) [social networks](/list/social-networks) [events](/list/events) **Social topic influence** [ai](/topic/ai), [open ai](/topic/open-ai), [applications](/topic/applications), [agi](/topic/agi), [winter](/topic/winter), [governance](/topic/governance), [model](/topic/model), [sama](/topic/sama), [paper](/topic/paper), [human](/topic/human) **Top assets mentioned** [Alphabet Inc Class A (GOOGL)](/topic/$googl) [Microsoft Corp. (MSFT)](/topic/microsoft) ### Top Social Posts Top posts by engagements in the last [--] hours "If we don't slow down generative AI prepare for: - Foreign states to steal base models they are years from building and fine-tune them as cyber weapons; - Mass voter manipulation and fake news without adequate safeguards. #pauseai" [X Link](https://x.com/anyuser/status/1660317499928494082) 2023-05-21T16:11Z [----] followers, [----] engagements "@LiamPaulGotch Sam Altman has discussed this in podcasts; I think Dario Amodei too" [X Link](https://x.com/ryan_kidd44/status/1714824785400906099) 2023-10-19T02:04Z [---] followers, [--] engagements "@pitdesi If you're founding a company to make money pick a board that makes money. If you're founding a company to steer the entire future pick a better board. They did" [X Link](https://x.com/ryan_kidd44/status/1726064830451491080) 2023-11-19T02:28Z [---] followers, [---] engagements "If OpenAI board fired @sama for straining charter but market forces put him back then Moloch wins" [X Link](https://x.com/anyuser/status/1726068314311627073) 2023-11-19T02:42Z [----] followers, 72.2K engagements "@NathanB60857242 @sama OpenAI charter seems like an attempt to stop Moloch from steering OpenAI. If charter can be subverted in Moloch's interests Moloch beats charter instigators" [X Link](https://x.com/ryan_kidd44/status/1726080446633779427) 2023-11-19T03:30Z [---] followers, [---] engagements "@OrionJohnston @QuintinPope5 @sama If you mean employees quitting I see your point. "Moloch" is a leaky abstraction anyways but I was using it principally to refer to investor pressure not employees quitting in solidarity" [X Link](https://x.com/ryan_kidd44/status/1726111607925903562) 2023-11-19T05:34Z [---] followers, [---] engagements "If @sama returns will it be because of: - Pressure from Microsoft; - OpenAI employees quitting in solidarity; - Board fears he will build a less safe competitor; - Unrelated board schism" [X Link](https://x.com/ryan_kidd44/status/1726141615558025238) 2023-11-19T07:33Z [---] followers, [---] engagements "@0xgokhan salam alaykum brother" [X Link](https://x.com/ryan_kidd44/status/1726361888773251115) 2023-11-19T22:08Z [---] followers, [--] engagements "I don't know @sama but I get the sense that: - Sam's love for OpenAI employees is sincere; - Sam cares about AI x-risk; - Sam thinks fast-deployment/slow-takeoff is optimally safe; - Sam would subvert the board for The Greater Good" [X Link](https://x.com/anyuser/status/1726364697862283278) 2023-11-19T22:19Z [----] followers, [----] engagements "This deserves way more attention. Zach built the best frontier AI lab safety scorecard on the internet evaluating @MicrosoftAI @GoogleDeepMind @OpenAI and @AnthropicAI I made an AI safety scorecard: I collected actions for frontier Al labs to avert extreme risks from AI then evaluated particular labs accordingly. https://t.co/4NsbT47BoL I made an AI safety scorecard: I collected actions for frontier Al labs to avert extreme risks from AI then evaluated particular labs accordingly. https://t.co/4NsbT47BoL" [X Link](https://x.com/ryan_kidd44/status/1785767701845250255) 2024-05-01T20:26Z [---] followers, [---] engagements "@kartographien @MicrosoftAI @GoogleDeepMind @OpenAI @AnthropicAI Zach is limited to public knowledge and objectively verifiable criteria. It's possible labs have much better internal governance structures and policies for the advent of AGI than the public knows" [X Link](https://x.com/ryan_kidd44/status/1785772703347318802) 2024-05-01T20:46Z [---] followers, [--] engagements "@kartographien @MicrosoftAI @GoogleDeepMind @OpenAI @AnthropicAI Also Zach gave substantial points to labs with governance structures that have no legal obligation towards shareholders + labs with whistleblower protections" [X Link](https://x.com/ryan_kidd44/status/1785772995480608980) 2024-05-01T20:47Z [---] followers, [--] engagements "The average mentor recommended MATS at 8.2/10 (NPS: +37) and improved their mentorship skills by 18%. Mentors thought within the next year 77% of scholars could achieve a first-author paper at a top conference 41% could receive a job offer from an AI lab safety team and 16% could found a new AI safety research organization" [X Link](https://x.com/ryan_kidd44/status/1789408301136122285) 2024-05-11T21:32Z [----] followers, [---] engagements "Many more scholars worked on interpretability than any other research area. Scholars were primarily interested in working at a scaling lab (Anthropic Google DeepMind OpenAI) or government organization (US/UK AISI). Mentors generally rated scholars' depth of technical ability and breadth of AI safety knowledge higher than self-reports" [X Link](https://x.com/ryan_kidd44/status/1789409443559690524) 2024-05-11T21:37Z [----] followers, [---] engagements "@The_JBernardi AI systems might be made inherently more interpretable than is possible with a human guaranteeing goal fidelity. Also they could submit proofs of their intent or otherwise engage in open source games with human overseers (who could communicate in code)" [X Link](https://x.com/ryan_kidd44/status/1810020648573678048) 2024-07-07T18:38Z [----] followers, [--] engagements "Saying that open weight AI models are the path to secure AI is like saying that sharing my psychological vulnerabilities with the world is the path to robust mental health" [X Link](https://x.com/anyuser/status/1821658217669390722) 2024-08-08T21:22Z [----] followers, [----] engagements "e/acc AGI realist humanist; pick two Nick Land says nothing human makes it out of the near-future and e/acc while being good PR is deluding itself to think otherwise https://t.co/CkGKUebhye Nick Land says nothing human makes it out of the near-future and e/acc while being good PR is deluding itself to think otherwise https://t.co/CkGKUebhye" [X Link](https://x.com/anyuser/status/1847692828656435618) 2024-10-19T17:34Z [----] followers, [----] engagements "@short_straw If you have long AGI timelines or are really optimistic it might be tempting to treat AI like any other technology in silicon valley and get mad at the government and doomers who are trying to slow down progress" [X Link](https://x.com/ryan_kidd44/status/1848158519692009514) 2024-10-21T00:25Z [----] followers, [--] engagements "@ChhatwalRaunak [--]. AGI might be here soon (2030). [--]. Time from AGI to superintelligence might be really quick (1-4 years). [--]. Superintelligence might be unsurvivable" [X Link](https://x.com/ryan_kidd44/status/1848205757944135804) 2024-10-21T03:32Z [----] followers, [--] engagements "OpenAI' latest model o3 scored: - [----] on Codeforces making it the 175th best competitive programmer on Earth - 25% on FrontierMath where "each problem demands hours of work from expert mathematicians" - 88% on GPQA where 70% represents PhD-level science knowledge - 88% on ARC-AGI where the average Mechanical Turk human worker scores 75% on hard visual reasoning problems" [X Link](https://x.com/anyuser/status/1870304098417066016) 2024-12-21T03:03Z [----] followers, 27.3K engagements "High-inference cost models like o3 might be a boon for AI safety: - More reasoning is done in chain-of-thought which is inspectable - Mech interp is more promising as base models will be smaller - Running frontier models will be more expensive reducing deployment overhang" [X Link](https://x.com/anyuser/status/1871303826260865433) 2024-12-23T21:16Z [----] followers, [----] engagements "The world is sleeping Survey of [---] experts by World Economic Forum reveals they have bizarre views about the biggest global risks. Most severe [--] year risk is extreme weather events https://t.co/0X16L0wFS1 Survey of [---] experts by World Economic Forum reveals they have bizarre views about the biggest global risks. Most severe [--] year risk is extreme weather events https://t.co/0X16L0wFS1" [X Link](https://x.com/anyuser/status/1885407492324417649) 2025-01-31T19:19Z [----] followers, 16.1K engagements "In defence of the first emotions often *are* weird hyper-specific reactions to internalized traumas or species-level herd instincts. But I think there's also deep meaning to many emotional triggers (e.g. self-other overlap) and outright rejecting emotional reactions seems a poor strategy for learning from them or constructing a robust moral framework grounded in universal instincts" [X Link](https://x.com/ryan_kidd44/status/1932669221290586216) 2025-06-11T05:20Z [----] followers, [---] engagements "Also without a good quantifiable definition of ethical delta (e.g. QALYs) it's pretty hard to hill-climb on improving the world; we're basically reduced to guesses and reading vibes or overly fixating on an easily optimized inadequate proxy like GDP" [X Link](https://x.com/ryan_kidd44/status/1932670526805463226) 2025-06-11T05:25Z [----] followers, [---] engagements "@robertskmiles Oof. I definitely have more mental machinery for bosons/fermions than for metaethics but I still think naming this thing has independent value" [X Link](https://x.com/ryan_kidd44/status/1939411387165573588) 2025-06-29T19:51Z [----] followers, [---] engagements "@mendezdotdev No you said "5% speed-up"" [X Link](https://x.com/ryan_kidd44/status/1943854776942571954) 2025-07-12T02:07Z [----] followers, [--] engagements "@BogdanIonutCir2 @nathan84686947 Speaking for MATS we're trying our best Usual sustainable growth rates are 1.25-1.5x per year; we've averaged 1.5x per year for the past three years (in terms of staff) and we 2x'd in 2025" [X Link](https://x.com/ryan_kidd44/status/1945566791394271378) 2025-07-16T19:30Z [----] followers, [--] engagements "@szarka @primalpoly @slow_developer @grok Even if trades are possible that doesn't mean humans get to have a living wage https://epoch.ai/gradient-updates/agi-could-drive-wages-below-subsistence-level https://epoch.ai/gradient-updates/agi-could-drive-wages-below-subsistence-level" [X Link](https://x.com/ryan_kidd44/status/1950939057519042844) 2025-07-31T15:18Z [----] followers, [--] engagements "80% of MATS alumni who completed the program before [----] are still working on AI safety today based on a survey of all available alumni LinkedIns or personal websites (242/292 83%). 10% are working on AI capabilities but only [--] on pre-training at a frontier AI company" [X Link](https://x.com/anyuser/status/1959721534886732260) 2025-08-24T20:56Z [----] followers, [----] engagements "Errata: I mistakenly included UK AISI in the "non-profit AI safety organization" category instead of "government agency"; updated plot below. I also mistakenly said that the [--] alumni working on AI capabilities at frontier AI companies were all working on pre-training" [X Link](https://x.com/ryan_kidd44/status/1960833659180892397) 2025-08-27T22:35Z [----] followers, [---] engagements "MATS [---] applications are open Launch your career in AI alignment governance and security with our 12-week research program. MATS provides field-leading research mentorship funding Berkeley & London offices housing and talks/workshops with AI experts" [X Link](https://x.com/anyuser/status/1961538891472916770) 2025-08-29T21:18Z [----] followers, 1M engagements "@AnthropicAI @GoogleDeepMind @OpenAI @AISecurityInst @RANDCorporation @redwood_ai @METR_Evals @apolloaievals 10% of MATS alumni who graduated before [----] co-founded active AI safety/security start-ups including @apolloaievals @Atla_AI @TimaeusResearch @Leap_Labs @theoremlabs @WorkshopLabsPBC and more" [X Link](https://x.com/anyuser/status/1961539137380782399) 2025-08-29T21:19Z [----] followers, [----] engagements "@AnthropicAI @GoogleDeepMind @OpenAI @AISecurityInst @RANDCorporation @redwood_ai @METR_Evals @apolloaievals @Atla_AI @TimaeusResearch @Leap_Labs @theoremlabs @WorkshopLabsPBC @sleepinyourhat @NeelNanda5 @EthanJPerez @McaleerStephen @vkrakovna @yonashav @StephenLCasper @bshlgrs @MariusHobbhahn @RichardMCNgo @janleike @AlecRad Apply by Oct [--] midnight AoE Visit the @MATSProgram website for detailed information on the program and application process. http://matsprogram.org/apply http://matsprogram.org/apply" [X Link](https://x.com/anyuser/status/1961540128553484684) 2025-08-29T21:23Z [----] followers, [----] engagements "@BogdanIonutCir2 @MATSprogram I dunno there are a lot of orgs here https://docs.google.com/spreadsheets/d/1moMMWNpi0PpvUMV8vMGpD5JouYR3OLUcKA_mo1tcu4Y/editusp=sharing https://docs.google.com/spreadsheets/d/1moMMWNpi0PpvUMV8vMGpD5JouYR3OLUcKA_mo1tcu4Y/editusp=sharing" [X Link](https://x.com/ryan_kidd44/status/1965122402121326639) 2025-09-08T18:37Z [----] followers, [--] engagements "@BogdanIonutCir2 @MATSprogram And tons of AI safety VCs: Babushkin Ventures Fifty Years Halcyon Futures Juniper Ventures Lionheart Ventures Macroscopic Ventures Metaplanet Mythos Ventures Safe Artificial Intelligence Fund" [X Link](https://x.com/ryan_kidd44/status/1965122567314047386) 2025-09-08T18:38Z [----] followers, [---] engagements "MATS Research Manager Jeffrey Heninger investigated the scaling rates of employees funding and revenue at frontier AI companies. https://www.lesswrong.com/posts/KW3nw5GYfnF9oNyp4/trends-in-economic-inputs-to-ai https://www.lesswrong.com/posts/KW3nw5GYfnF9oNyp4/trends-in-economic-inputs-to-ai" [X Link](https://x.com/ryan_kidd44/status/1966344733774721279) 2025-09-12T03:34Z [----] followers, [----] engagements "The growth rate for capital at OpenAI is 2.2x per year at Anthropic is 3.5x per year and at xAI is 3.3x per year" [X Link](https://x.com/ryan_kidd44/status/1966346809451884656) 2025-09-12T03:43Z [----] followers, [---] engagements "The growth rate in revenue for OpenAI is 3.2x per year. Other frontier AI companies have unclear revenue data" [X Link](https://x.com/ryan_kidd44/status/1966347491856859580) 2025-09-12T03:45Z [----] followers, [---] engagements "The AI safety & security research field is growing by 25% per year. At this rate there will be 8.5k researchers when we reach AGI" [X Link](https://x.com/ryan_kidd44/status/1972384232598294563) 2025-09-28T19:33Z [----] followers, 47.5K engagements "Data from Stephen McAleese: AI milestones from Metaculus: https://www.metaculus.com/questions/19356/transformative-ai-date/ https://www.metaculus.com/questions/5121/date-of-artificial-general-intelligence/ https://www.metaculus.com/questions/3479/date-weakly-general-ai-is-publicly-known/ https://www.lesswrong.com/posts/8QjAnWyuE9fktPRgS/ai-safety-field-growth-analysis-2025 https://www.metaculus.com/questions/19356/transformative-ai-date/ https://www.metaculus.com/questions/5121/date-of-artificial-general-intelligence/" [X Link](https://x.com/ryan_kidd44/status/1972384775773340048) 2025-09-28T19:35Z [----] followers, [---] engagements "@provisionalidea Actually we focus a lot on AI security control and misuse prevention at MATS. [---] AI safety and security researchers (and [---] AI gov) might seem like a lot but it's puny compared to the OS community and we still get zero days AI security is plain hard and needs talent" [X Link](https://x.com/ryan_kidd44/status/1972695342610358404) 2025-09-29T16:09Z [----] followers, [--] engagements "@dromanocpm What about for people who would be stay in academia regardless for whom mech interp is just a better use of their time" [X Link](https://x.com/ryan_kidd44/status/1670212309183250435) 2023-06-17T23:30Z [----] followers, [--] engagements "Is "superconsciousness" possible How would it differ from regular consciousness albeit run on vastly more expensive hardware" [X Link](https://x.com/ryan_kidd44/status/1931895597814186069) 2025-06-09T02:06Z [----] followers, [---] engagements "@edzitron I assume Anthropic is spending most of their revenue on R&D due to the AGI race https://epoch.ai/data-insights/openai-compute-spend https://epoch.ai/data-insights/openai-compute-spend" [X Link](https://x.com/ryan_kidd44/status/1980409403426894059) 2025-10-20T23:02Z [----] followers, [----] engagements "What is the chief bottleneck in scaling the AI safety field My hypothesis: research program/people managers. MATS applicants are increasing exponentially and we had [---] mentors apply for Summer [----] (and expect to accept 20%). Great research managers are our chief constraint" [X Link](https://x.com/ryan_kidd44/status/1982862838747279614) 2025-10-27T17:31Z [----] followers, [---] engagements "@eigen_spectra @yonashav @sebkrier Re. field-builders former MATS Research Managers have gone on to (re)found Constellation's Astra Fellowship manage Anthropic's AI safety external partnerships and help other AI safety orgs scale. I think growing our RM pipeline is a good way to grow field-builders" [X Link](https://x.com/ryan_kidd44/status/1982892569756676539) 2025-10-27T19:30Z [----] followers, [--] engagements "@eigen_spectra @yonashav @sebkrier Re. founders we are seeing some great early results from @HalcyonFutures @catalyzeimpact @seldonai. Future of Life Foundation Fifty Years [----] Atlas Computing and more are entering the fray" [X Link](https://x.com/ryan_kidd44/status/1982893290791039474) 2025-10-27T19:32Z [----] followers, [--] engagements "I wrote a blog post on why I think the AI safety ecosystem undervalues founders and field-builders and what to do about it https://www.lesswrong.com/posts/yw9B5jQazBKGLjize/ai-safety-undervalues-founders https://www.lesswrong.com/posts/yw9B5jQazBKGLjize/ai-safety-undervalues-founders" [X Link](https://x.com/ryan_kidd44/status/1989905011770622447) 2025-11-16T03:55Z [----] followers, [----] engagements "Hackathon for AI safety Online from 7:30 pm today and in SF all day tomorrow. DEF/ACC HACKATHON NOV [--] @ VIVARIUM SF WHAT ARE WE ACCELERATING TOWARDS https://t.co/x1r3mr2hFF DEF/ACC HACKATHON NOV [--] @ VIVARIUM SF WHAT ARE WE ACCELERATING TOWARDS https://t.co/x1r3mr2hFF" [X Link](https://x.com/ryan_kidd44/status/1991957684229419168) 2025-11-21T19:51Z [----] followers, [---] engagements "Interested in raising awareness for AI safety via digital content Apply to the Frame Fellowship Mentors include @robertskmiles @AricFloyd @PauseusMaximus https://framefellowship.com/ https://framefellowship.com/" [X Link](https://x.com/ryan_kidd44/status/1993111893301772436) 2025-11-25T00:18Z [----] followers, [---] engagements "NeurIPS [----] was an amazing time for @MATSprogram Our fellows presented [--] spotlight papers (top 2%) we received over [---] sign-ups from our sponsor booth and many fellows presented papers at the Mech Interp and Alignment Workshops" [X Link](https://x.com/ryan_kidd44/status/1999599413061386578) 2025-12-12T21:57Z [----] followers, [----] engagements "Spotlit papers: - Distillation Robustifies Unlearning - Among Us: A Sandbox for Measuring and Detecting Agentic Deception - Overcoming Sparsity Artifacts in Crosscoders to Interpret Chat-Tuning - SAGE-Eval: Evaluating LLMs for Systematic Generalizations of Safety Facts - The Non-Linear Representation Dilemma: Is Causal Abstraction Enough for Mechanistic Interpretability https://arxiv.org/abs/2507.08802 https://arxiv.org/abs/2505.21828 https://arxiv.org/abs/2504.02922 https://arxiv.org/abs/2504.04072 https://arxiv.org/abs/2506.06278 https://arxiv.org/abs/2504.02922" [X Link](https://x.com/ryan_kidd44/status/1999600667414114471) 2025-12-12T22:02Z [----] followers, [----] engagements "@gmiller @TransluceAI More of a cute example than anything you can base strategy around but I like Wei Dai's game where the smartest players lose: http://www.weidai.com/smart-losers.txt http://www.weidai.com/smart-losers.txt" [X Link](https://x.com/ryan_kidd44/status/2001733453390451180) 2025-12-18T19:17Z [----] followers, [--] engagements "What is most blocking talent from contributing to AI safety & security Rare skills proof of competence and personal connections" [X Link](https://x.com/ryan_kidd44/status/2023108374406332803) 2026-02-15T18:53Z [----] followers, [----] engagements "Becoming an effective AI safety researcher typically requires experienced mentorship peer feedback and many repetitions of working on real problems particularly for building research taste. Self study is often not enough and experienced mentors are a bottleneck" [X Link](https://x.com/ryan_kidd44/status/2023109147726840173) 2026-02-15T18:56Z [----] followers, [---] engagements "@MATSprogram Winter 2024-25 mentors include researchers from @AnthropicAI @GoogleDeepMind @AISafetyInst @CNASdc @CHAI_Berkeley @AlgAlignMIT @farairesearch @cais @apolloaisafety @kasl_ai @MIRIBerkeley and more Apply by Oct [--]. https://www.matsprogram.org/mentors http://redwoodresearch.org https://www.matsprogram.org/mentors http://redwoodresearch.org" [X Link](https://x.com/anyuser/status/1834346185341239713) 2024-09-12T21:39Z [----] followers, 35.6K engagements "MATS received a record-breaking [----] applications for our Summer program and plan to accept 5%. We also received [---] mentor applications and accepted 20% as primary mentors. Applicants are growing exponentially at 2x/year" [X Link](https://x.com/ryan_kidd44/status/2015911887993757884) 2026-01-26T22:17Z [----] followers, 18.4K engagements "@beyarkay Here is a plot of annual citations from the only three AI safety nonprofits with Google Scholar pages. https://scholar.google.com/citationsuser=VgJaUK4AAAAJ&hl=en https://scholar.google.com/citationsuser=VgJaUK4AAAAJ&hl=en" [X Link](https://x.com/ryan_kidd44/status/2016915499347841238) 2026-01-29T16:45Z [----] followers, [---] engagements "I had a great time chatting with Jacob Haines about AI safety field-building and emerging talent needs https://kairos.fm/intoaisafety/e027/ https://kairos.fm/intoaisafety/e027/" [X Link](https://x.com/ryan_kidd44/status/2018828801158025272) 2026-02-03T23:27Z [----] followers, [---] engagements "@hamsabastani @TomCunningham75 @joel_bkr https://arxiv.org/abs/2109.08065 https://arxiv.org/abs/2109.08065" [X Link](https://x.com/ryan_kidd44/status/2019809512254062987) 2026-02-06T16:24Z [----] followers, [---] engagements "What proportion of ML academics are interested in AI safety I analyzed the research interests of the [---] Action Editors on TMLR Editorial Board. 4% are interested in alignment or safety; 10% if you include interp evals trust or security. https://jmlr.org/tmlr/editorial-board.html https://jmlr.org/tmlr/editorial-board.html" [X Link](https://x.com/ryan_kidd44/status/2020190277307347016) 2026-02-07T17:37Z [----] followers, [----] engagements "@ExcelEthicsAI Actually it's pretty cheap to say I'm interested in AI safety but time-expensive to actually publish on the subject. I am not convinced these statistics reflect publication patterns rather than sentiment" [X Link](https://x.com/ryan_kidd44/status/2020192666580054203) 2026-02-07T17:47Z [----] followers, [--] engagements "AI safety field-building in Australia should accelerate. OpenAI and Anthropic opened Sydney offices OpenAI started building a $4.6B datacenter in Sydney and the country is a close US/UK ally. https://www.lesswrong.com/posts/tPjAgWpsQrveFECWP/ryan-kidd-s-shortformcommentId=aPAtazuRt2np2zn6n https://www.lesswrong.com/posts/tPjAgWpsQrveFECWP/ryan-kidd-s-shortformcommentId=aPAtazuRt2np2zn6n" [X Link](https://x.com/ryan_kidd44/status/2020292079902159316) 2026-02-08T00:22Z [----] followers, [----] engagements "MATS [----] applications are open Launch your career in AI alignment governance and security with our 12-week research program. MATS provides field-leading research mentorship funding Berkeley & London offices housing and talks/workshops with AI experts" [X Link](https://x.com/ryan_kidd44/status/2001005525811769454) 2025-12-16T19:04Z [----] followers, 2.2M engagements "Aspiring researchers need a portfolio of research outputs references from credible supervisors and credentials that signal competence to potential employers and funders (e.g. MATS and BlueDot on their CV). Without these even talented individuals miss opportunities" [X Link](https://x.com/ryan_kidd44/status/2023109468230394301) 2026-02-15T18:57Z [----] followers, [---] engagements "If you're hiring in AI alignment interpretability governance or security MATS can help We maintain a database of alumni looking for work and can make targeted headhunting recommendations" [X Link](https://x.com/anyuser/status/2017056474372198672) 2026-01-30T02:05Z [----] followers, [----] engagements "If OpenAI board fired @sama for straining charter but market forces put him back then Moloch wins" [X Link](https://x.com/anyuser/status/1726068314311627073) 2023-11-19T02:42Z [----] followers, 72.2K engagements "MATS [---] applications are open Launch your career in AI alignment governance and security with our 12-week research program. MATS provides field-leading research mentorship funding Berkeley & London offices housing and talks/workshops with AI experts" [X Link](https://x.com/anyuser/status/1961538891472916770) 2025-08-29T21:18Z [----] followers, 1M engagements "Summer applications just launched Mentors include AI safety researchers from @AnthropicAI @OpenAI @deepmind @MIRIBerkeley @CHAI_Berkeley @cais @FHIOxford and more https://www.serimats.org/ https://www.serimats.org/" [X Link](https://x.com/anyuser/status/1644817241664536578) 2023-04-08T21:39Z [----] followers, 34K engagements "The world is sleeping Survey of [---] experts by World Economic Forum reveals they have bizarre views about the biggest global risks. Most severe [--] year risk is extreme weather events https://t.co/0X16L0wFS1 Survey of [---] experts by World Economic Forum reveals they have bizarre views about the biggest global risks. Most severe [--] year risk is extreme weather events https://t.co/0X16L0wFS1" [X Link](https://x.com/anyuser/status/1885407492324417649) 2025-01-31T19:19Z [----] followers, 16.1K engagements "Survey of [---] experts by World Economic Forum reveals they have bizarre views about the biggest global risks. Most severe [--] year risk is extreme weather events" [X Link](https://x.com/anyuser/status/1885107313616273878) 2025-01-30T23:26Z 16.8K followers, 413.7K engagements "80% of MATS alumni who completed the program before [----] are still working on AI safety today based on a survey of all available alumni LinkedIns or personal websites (242/292 83%). 10% are working on AI capabilities but only [--] on pre-training at a frontier AI company" [X Link](https://x.com/anyuser/status/1959721534886732260) 2025-08-24T20:56Z [----] followers, [----] engagements "In [---] years @MATSprogram has helped produce [---] arXiv publications. Our organizational h-index is 31" [X Link](https://x.com/anyuser/status/1950597247424668129) 2025-07-30T16:39Z [----] followers, [----] engagements "@MATSprogram Summer [----] applications close Apr [--] Come help advance the fields of AI alignment security and governance with mentors including @NeelNanda5 @EthanJPerez @OwainEvans_UK @EvanHub @bshlgrs @dawnsongtweets @DavidSKrueger @RichardMCNgo and more" [X Link](https://x.com/anyuser/status/1902548131973353512) 2025-03-20T02:29Z [----] followers, 57.9K engagements "OpenAI' latest model o3 scored: - [----] on Codeforces making it the 175th best competitive programmer on Earth - 25% on FrontierMath where "each problem demands hours of work from expert mathematicians" - 88% on GPQA where 70% represents PhD-level science knowledge - 88% on ARC-AGI where the average Mechanical Turk human worker scores 75% on hard visual reasoning problems" [X Link](https://x.com/anyuser/status/1870304098417066016) 2024-12-21T03:03Z [----] followers, 27.3K engagements "MATS has received mentorship applications from [---] researchers for our Winter [----] program far more than we can support. If you run an AI safety or governance program and you want referrals let me know" [X Link](https://x.com/anyuser/status/1929675104163975468) 2025-06-02T23:02Z [----] followers, [----] engagements "What a week Both Anthropic and DeepMind shed some light on their AI alignment plans after OpenAI shared their plan in Aug [----]. * Anthropic: * DeepMind: * OpenAI: https://openai.com/blog/our-approach-to-alignment-research https://www.lesswrong.com/./4iEpGXb./p/a9SPcZ6GXAg9cNKdi https://www.anthropic.com/index/core-views-on-ai-safety https://openai.com/blog/our-approach-to-alignment-research https://www.lesswrong.com/./4iEpGXb./p/a9SPcZ6GXAg9cNKdi https://www.anthropic.com/index/core-views-on-ai-safety" [X Link](https://x.com/anyuser/status/1633954828824252417) 2023-03-09T22:16Z [----] followers, [----] engagements "The Paris AI Summit was a staggering failure. Entropy - [--] Humanity - [--] https://www.transformernews.ai/p/paris-ai-summit-failure https://www.transformernews.ai/p/paris-ai-summit-failure" [X Link](https://x.com/anyuser/status/1889498041402663332) 2025-02-12T02:13Z [----] followers, 12.4K engagements "First AI alignment paper to win ICML Best Paper So happy to have helped support this work at @MATSprogram :) Well done @McHughes288 @danvalentine256 @sleight_henry @akbirkhan @EthanJPerez @sleepinyourhat and coauthors excited to announce this received an ICML Best Paper Award come see our talk at 10:30 tomorrow https://t.co/PCH1q0f0Po excited to announce this received an ICML Best Paper Award come see our talk at 10:30 tomorrow https://t.co/PCH1q0f0Po" [X Link](https://x.com/anyuser/status/1815845406645039554) 2024-07-23T20:24Z [----] followers, 11.4K engagements "excited to announce this received an ICML Best Paper Award come see our talk at 10:30 tomorrow How can we check LLM outputs in domains where we are not experts We find that non-expert humans answer questions better after reading debates between expert LLMs. Moreover human judges are more accurate as experts get more persuasive. 📈 https://t.co/jgyfCEQvfw https://t.co/wWRWxojD6H How can we check LLM outputs in domains where we are not experts We find that non-expert humans answer questions better after reading debates between expert LLMs. Moreover human judges are more accurate as experts get" [X Link](https://x.com/anyuser/status/1815486293692019178) 2024-07-22T20:37Z [----] followers, 76.6K engagements "Over [----] applicants to @MATSprogram; what a milestone 🎉" [X Link](https://x.com/anyuser/status/1776487781549605363) 2024-04-06T05:51Z [----] followers, [----] engagements "Another @MATSprogram concluded and another [--] scholars graduated on Fri Our Scholar Symposium featured [--] talks on AI interpretability model evals + demos agent foundations control/red-teaming scalable oversight and more" [X Link](https://x.com/anyuser/status/1769481336690020818) 2024-03-17T21:50Z [----] followers, [----] engagements "Reasons to be optimistic about AI x-safety: [--]. The public cares more than expected; [--]. Governments aren't ignoring the problem; [--]. LMs might be much more interpretable than end-to-end RL; [--]. Instructed LMs might generalize better than expected" [X Link](https://x.com/anyuser/status/1714413321590550585) 2023-10-17T22:49Z [----] followers, [----] engagements "Three AI safety boys in Tokyo" [X Link](https://x.com/anyuser/status/1777294117665640712) 2024-04-08T11:15Z [----] followers, [----] engagements "e/acc AGI realist humanist; pick two Nick Land says nothing human makes it out of the near-future and e/acc while being good PR is deluding itself to think otherwise https://t.co/CkGKUebhye Nick Land says nothing human makes it out of the near-future and e/acc while being good PR is deluding itself to think otherwise https://t.co/CkGKUebhye" [X Link](https://x.com/anyuser/status/1847692828656435618) 2024-10-19T17:34Z [----] followers, [----] engagements "Nick Land says nothing human makes it out of the near-future and e/acc while being good PR is deluding itself to think otherwise" [X Link](https://x.com/anyuser/status/1845579693816438987) 2024-10-13T21:37Z 84.6K followers, 74.3K engagements "Who are the top ethicists working on: - What values to instill in artificial superintelligence - How should AI-generated wealth be distributed - What should people do in a post-labor society - What level of surveillance/restriction is justified by the Unilateralist's Curse" [X Link](https://x.com/anyuser/status/1809713187484479618) 2024-07-06T22:17Z [----] followers, [----] engagements "I think a lot of mechanistic interpretability research should find a home in academic labs because: [--]. Mech interp isn't very expensive; [--]. Related academic research (e.g. sparsity pruning) is strong; [--]. Mech interp should grow; [--]. Most academic safety research is less useful" [X Link](https://x.com/anyuser/status/1670189227626356737) 2023-06-17T21:58Z [----] followers, [----] engagements "The MATS Winter 2023-24 Cohort has launched Apply by Nov [--] to help advance AI safety. (Note: Neel Nanda's applications close early on Nov 10) https://www.matsprogram.org/ https://www.matsprogram.org/" [X Link](https://x.com/anyuser/status/1715538860980179274) 2023-10-21T01:21Z [----] followers, 20.5K engagements "I pre-ordered this and you should too https://ifanyonebuildsit.com https://ifanyonebuildsit.com" [X Link](https://x.com/anyuser/status/1936930391467966926) 2025-06-22T23:32Z [----] followers, [----] engagements "Applications are open for @MATSprogram Summer [----] (Jun 17-Aug 23) and Winter [----] (Jan 6-Mar 14) Deadline is Mar [--]. Apply here (10 min) https://airtable.com/appPxJ0QMqR7TElYU/pagRPwHQtcN8L0vIE/form https://airtable.com/appPxJ0QMqR7TElYU/pagRPwHQtcN8L0vIE/form" [X Link](https://x.com/anyuser/status/1766230485657850228) 2024-03-08T22:32Z [----] followers, 28.3K engagements "Saying that open weight AI models are the path to secure AI is like saying that sharing my psychological vulnerabilities with the world is the path to robust mental health" [X Link](https://x.com/anyuser/status/1821658217669390722) 2024-08-08T21:22Z [----] followers, [----] engagements "@MATSprogram has [----] summer applicants and enough funding to accept 2.5% (ideally 7%). Accepting donations via and at $24k/scholar. Help us support mentors like @NeelNanda5 @OwainEvans_UK @EthanJPerez @EvanHub and more http://manifund.org/projects/mats-funding http://existence.org http://manifund.org/projects/mats-funding http://existence.org" [X Link](https://x.com/anyuser/status/1784374612555641296) 2024-04-28T00:10Z [----] followers, [----] engagements "@MATSprogram Winter 2024-25 mentors include researchers from @AnthropicAI @GoogleDeepMind @AISafetyInst @CNASdc @CHAI_Berkeley @AlgAlignMIT @farairesearch @cais @apolloaisafety @kasl_ai @MIRIBerkeley and more Apply by Oct [--]. https://www.matsprogram.org/mentors http://redwoodresearch.org https://www.matsprogram.org/mentors http://redwoodresearch.org" [X Link](https://x.com/anyuser/status/1834346185341239713) 2024-09-12T21:39Z [----] followers, 35.6K engagements "Situational Awareness [---] "How exactly could AI take over by 2027" Introducing AI 2027: a deeply-researched scenario forecast I wrote alongside @slatestarcodex @eli_lifland and @thlarsen https://t.co/v0V0RbFoVA "How exactly could AI take over by 2027" Introducing AI 2027: a deeply-researched scenario forecast I wrote alongside @slatestarcodex @eli_lifland and @thlarsen https://t.co/v0V0RbFoVA" [X Link](https://x.com/anyuser/status/1907830901284941933) 2025-04-03T16:21Z [----] followers, [----] engagements ""How exactly could AI take over by 2027" Introducing AI 2027: a deeply-researched scenario forecast I wrote alongside @slatestarcodex @eli_lifland and @thlarsen" [X Link](https://x.com/anyuser/status/1907826614186209524) 2025-04-03T16:04Z 29.6K followers, 2.9M engagements "High-inference cost models like o3 might be a boon for AI safety: - More reasoning is done in chain-of-thought which is inspectable - Mech interp is more promising as base models will be smaller - Running frontier models will be more expensive reducing deployment overhang" [X Link](https://x.com/anyuser/status/1871303826260865433) 2024-12-23T21:16Z [----] followers, [----] engagements "An enigma a shoggoth and two ex-physicists walk into a party" [X Link](https://x.com/anyuser/status/1634288077421318144) 2023-03-10T20:20Z [----] followers, [----] engagements "Technical AI alignment/control is still impactful; don't go all-in on AI gov - Liability incentivises safeguards even absent regulation; - Cheaper more effective safeguards make it easier for labs to meet safety standards; - Concrete safeguards give regulation teeth" [X Link](https://x.com/anyuser/status/1935405219040657509) 2025-06-18T18:32Z [----] followers, [----] engagements "So proud of all of these MATS scholars and their projects https://drive.google.com/file/d/1HA5RUCM15-6COISmdkGF2w_JmCMzlQNy/viewusp=drivesdk https://drive.google.com/file/d/1HA5RUCM15-6COISmdkGF2w_JmCMzlQNy/viewusp=drivesdk" [X Link](https://x.com/anyuser/status/1698758122230419839) 2023-09-04T18:01Z [----] followers, [----] engagements "MATS Winter 2024-25 applications close Oct [--] Come and kick-start your AI safety research career. Mentors include @OwainEvans_UK @bshlgrs @EvanHub @StephenLCasper and more https://matsprogram.org https://matsprogram.org" [X Link](https://x.com/anyuser/status/1834657755090895115) 2024-09-13T18:17Z [----] followers, [----] engagements "If we don't slow down generative AI prepare for: - Foreign states to steal base models they are years from building and fine-tune them as cyber weapons; - Mass voter manipulation and fake news without adequate safeguards. #pauseai" [X Link](https://x.com/anyuser/status/1660317499928494082) 2023-05-21T16:11Z [----] followers, [----] engagements "Excited to present at the Technical AI Safety Conference in Tokyo https://tais2024.cc/ https://tais2024.cc/" [X Link](https://x.com/anyuser/status/1772745449461735689) 2024-03-26T22:00Z [----] followers, [----] engagements "This deserves way more attention. Zach built the best frontier AI lab safety scorecard on the internet evaluating @MicrosoftAI @GoogleDeepMind @AIatMeta @GoogleDeepMind and @AnthropicAI I made an AI safety scorecard: I collected actions for frontier Al labs to avert extreme risks from AI then evaluated particular labs accordingly. https://t.co/4NsbT47BoL I made an AI safety scorecard: I collected actions for frontier Al labs to avert extreme risks from AI then evaluated particular labs accordingly. https://t.co/4NsbT47BoL" [X Link](https://x.com/anyuser/status/1785773101424562521) 2024-05-01T20:47Z [----] followers, 11.1K engagements "I made an AI safety scorecard: I collected actions for frontier Al labs to avert extreme risks from AI then evaluated particular labs accordingly. http://ailabwatch.org http://ailabwatch.org" [X Link](https://x.com/anyuser/status/1785353610421002657) 2024-04-30T17:00Z [--] followers, [----] engagements "The OpenAI plan seems to have been: Pragmatism: Lead the AI pack from the front instead of someone worse. Caution: Keep risk within tolerance or pull the plug. If this is "pulling the plug" and it fails I am pessimistic about all such plans" [X Link](https://x.com/anyuser/status/1726074227776950284) 2023-11-19T03:05Z [----] followers, [----] engagements "@MATSprogram 10% of our [---] alumni have co-founded organizations or research teams during or after MATS" [X Link](https://x.com/anyuser/status/1950600411314008080) 2025-07-30T16:52Z [----] followers, [---] engagements "Now hiring ops generalists community manager and research coaches to grow AI safety https://tinyurl.com/2v7pad7u https://tinyurl.com/2v7pad7u" [X Link](https://x.com/anyuser/status/1661086871081070596) 2023-05-23T19:09Z [----] followers, [----] engagements "EAs seem to come in two primary flavors: - Specialists with high cognitive empathy who want to make utilons go up; - Generalists with high affective empathy who want to empower all beings" [X Link](https://x.com/anyuser/status/1932600996292931697) 2025-06-11T00:49Z [----] followers, [----] engagements "Anthropic Fellows has produced some awesome AI safety research Were hiring someone to run the Anthropic Fellows Program Our research collaborations have led to some of our best safety research and hires. Were looking for an exceptional ops generalist TPM or research/eng manager to help us significantly scale and improve our collabs 🧵 Were hiring someone to run the Anthropic Fellows Program Our research collaborations have led to some of our best safety research and hires. Were looking for an exceptional ops generalist TPM or research/eng manager to help us significantly scale and improve our" [X Link](https://x.com/anyuser/status/1963829776625631468) 2025-09-05T05:01Z [----] followers, [----] engagements "Were hiring someone to run the Anthropic Fellows Program Our research collaborations have led to some of our best safety research and hires. Were looking for an exceptional ops generalist TPM or research/eng manager to help us significantly scale and improve our collabs 🧵" [X Link](https://x.com/anyuser/status/1963664611397546145) 2025-09-04T18:05Z 13.4K followers, 68.4K engagements "Amazing work Marius So happy to have helped support Apollo and your journey via @MATSprogram Honored and humbled to be in @TIME's list of the TIME100 AI of [----] https://t.co/mz17wPSuaL #TIME100AI https://t.co/uDvdCYvE9j Honored and humbled to be in @TIME's list of the TIME100 AI of [----] https://t.co/mz17wPSuaL #TIME100AI https://t.co/uDvdCYvE9j" [X Link](https://x.com/anyuser/status/1961124949239718002) 2025-08-28T17:53Z [----] followers, [---] engagements "Honored and humbled to be in @TIME's list of the TIME100 AI of [----] #TIME100AI https://time.com/collections/time100-ai-2025/7305864/marius-hobbhahn/ https://time.com/collections/time100-ai-2025/7305864/marius-hobbhahn/" [X Link](https://x.com/anyuser/status/1961046074262512041) 2025-08-28T12:39Z [----] followers, 17.4K engagements "Applications to @NeelNanda5's mech interp @MATSprogram are now open Apply by Aug [--]. https://forms.matsprogram.org/general-application Are you excited about @ch402-style mechanistic interpretability research I'm looking to mentor scholars via MATS - apply by Aug [--] I'm impressed by the work from past scholars and love mentoring promising talent. You don't need to be in a big lab to do good mech interp work https://forms.matsprogram.org/general-application Are you excited about @ch402-style mechanistic interpretability research I'm looking to mentor scholars via MATS - apply by Aug [--] I'm" [X Link](https://x.com/anyuser/status/1815471824035348793) 2024-07-22T19:39Z [----] followers, [----] engagements "Are you excited about @ch402-style mechanistic interpretability research I'm looking to mentor scholars via MATS - apply by Aug [--] I'm impressed by the work from past scholars and love mentoring promising talent. You don't need to be in a big lab to do good mech interp work" [X Link](https://x.com/anyuser/status/1813921161052635209) 2024-07-18T12:58Z 36.4K followers, 46.8K engagements "I don't know @sama but I get the sense that: - Sam's love for OpenAI employees is sincere; - Sam cares about AI x-risk; - Sam thinks fast-deployment/slow-takeoff is optimally safe; - Sam would subvert the board for The Greater Good" [X Link](https://x.com/anyuser/status/1726364697862283278) 2023-11-19T22:19Z [----] followers, [----] engagements "Reasons to be pessimistic about AI x-safety: [--]. We might have less time than we thought; [--]. The current best plan relies on big tech displaying a vastly better security mindset than usual; [--]. There seems to be a shortage of new good ideas for AI alignment; [--]. A few actors (e.g. SBF) might have harmed the public image of orgs/movements pushing for AI x-safety" [X Link](https://x.com/anyuser/status/1714437105320173702) 2023-10-18T00:23Z [----] followers, [---] engagements "Very happy to have supported this research at @MATSprogram. Applications for Summer [----] launching soon New Anthropic research: Auditing Language Models for Hidden Objectives. We deliberately trained a model with a hidden misaligned objective and put researchers to the test: Could they figure out the objective without being told https://t.co/fxmA9Os2C9 New Anthropic research: Auditing Language Models for Hidden Objectives. We deliberately trained a model with a hidden misaligned objective and put researchers to the test: Could they figure out the objective without being told" [X Link](https://x.com/anyuser/status/1900295898799980581) 2025-03-13T21:20Z [----] followers, [---] engagements "New Anthropic research: Auditing Language Models for Hidden Objectives. We deliberately trained a model with a hidden misaligned objective and put researchers to the test: Could they figure out the objective without being told" [X Link](https://x.com/anyuser/status/1900217234825634236) 2025-03-13T16:07Z 837.8K followers, 354.9K engagements "@AnthropicAI @GoogleDeepMind @OpenAI @AISecurityInst @RANDCorporation @redwood_ai @METR_Evals @apolloaievals @Atla_AI @TimaeusResearch @Leap_Labs @theoremlabs @WorkshopLabsPBC MATS connects researchers with world-class mentors including @sleepinyourhat Nicholas Carlini @NeelNanda5 @EthanJPerez @McaleerStephen @vkrakovna @yonashav @StephenLCasper @bshlgrs @MariusHobbhahn @RichardMCNgo and more @janleike @AlecRad etc. often collaborate as advisors" [X Link](https://x.com/anyuser/status/1961539983300596017) 2025-08-29T21:22Z [----] followers, [----] engagements "@AnthropicAI @GoogleDeepMind @OpenAI @AISecurityInst @RANDCorporation @redwood_ai @METR_Evals @apolloaievals @Atla_AI @TimaeusResearch @Leap_Labs @theoremlabs @WorkshopLabsPBC @sleepinyourhat @NeelNanda5 @EthanJPerez @McaleerStephen @vkrakovna @yonashav @StephenLCasper @bshlgrs @MariusHobbhahn @RichardMCNgo @janleike @AlecRad Apply by Oct [--] midnight AoE Visit the @MATSProgram website for detailed information on the program and application process. http://matsprogram.org/apply http://matsprogram.org/apply" [X Link](https://x.com/anyuser/status/1961540128553484684) 2025-08-29T21:23Z [----] followers, [----] engagements "Currently accepting AI safety research mentors for a Winter program; message me if you are interested Past mentors include: http://serimats.org/mentors http://serimats.org/mentors" [X Link](https://x.com/anyuser/status/1699197088754765914) 2023-09-05T23:05Z [----] followers, [----] engagements "Also: - What moral personhood will digital minds have - How should nations share decision making power regarding world-transforming and Mercury-disassembling technology" [X Link](https://x.com/anyuser/status/1809746460415795257) 2024-07-07T00:29Z [----] followers, [----] engagements "Big support to @austinc3301 and his new project AI safety student groups and entry-level internships are very important to the @MATSprogram pipeline (and all of AI safety). On more personal news I'm now the Co-Director of Kairos a new AI safety fieldbuilding org https://t.co/GAsLMmfSlY On more personal news I'm now the Co-Director of Kairos a new AI safety fieldbuilding org https://t.co/GAsLMmfSlY" [X Link](https://x.com/anyuser/status/1849952526839447644) 2024-10-25T23:13Z [----] followers, [----] engagements "On more personal news I'm now the Co-Director of Kairos a new AI safety fieldbuilding org https://forum.effectivealtruism.org/posts/tTCyWthLyHk32WX9n/introducing-kairos-a-new-home-for-spar-and-fsp https://forum.effectivealtruism.org/posts/tTCyWthLyHk32WX9n/introducing-kairos-a-new-home-for-spar-and-fsp" [X Link](https://x.com/anyuser/status/1849936485497086186) 2024-10-25T22:10Z [----] followers, [----] engagements "AI safety research that reduces the risk of non-catastrophic accidents or misuse (e.g. hate speech) makes commercial AI more viable driving AI hype and capabilities research. While important this research might fail to prevent genuinely catastrophic "black swan" risk" [X Link](https://x.com/anyuser/status/1635697142097379349) 2023-03-14T17:39Z [----] followers, [----] engagements "Ever wanted to contribute to technical AI safety but haven't built a transformer Apply to The ML for AI safety bootcamp will run Sep 2-Oct [--] out of Applications close Jul [--]. http://SafeAI.org.uk http://ARENA.education http://SafeAI.org.uk http://ARENA.education" [X Link](https://x.com/anyuser/status/1810760612672327923) 2024-07-09T19:39Z [----] followers, [---] engagements "@mealreplacer Good evening Robert" [X Link](https://x.com/anyuser/status/1661118131236904995) 2023-05-23T21:13Z [----] followers, [---] engagements "MATS has accelerated 450+ researchers in the past [---] years. 80% of MATS alumni who graduated before [----] are working on AI safety/security including 200+ at @AnthropicAI @GoogleDeepMind @OpenAI @AISecurityInst @RANDCorporation @redwood_ai @METR_Evals @apolloaievals and more" [X Link](https://x.com/anyuser/status/1961539040383308113) 2025-08-29T21:18Z [----] followers, [----] engagements "Our selection process is ongoing but it looks like around 75% of mentor applicants are rated above our minimum bar by our Mentor Selection Committee. If we accept [--] mentors that leaves [--] great mentors unsupported" [X Link](https://x.com/anyuser/status/1929675852872405294) 2025-06-02T23:05Z [----] followers, [----] engagements "Precipissed: Feeling angry about civilizational inadequacy towards mitigating x-risk" [X Link](https://x.com/anyuser/status/1603483541131497472) 2022-12-15T20:14Z [----] followers, [---] engagements "Excited to have supported this research @MATSprogram New paper on evaluating instrumental self-reasoning ability in frontier models 🤖🪞 We propose a suite of agentic tasks that are more diverse than prior work and give us a more representative picture of how good models are at eg. self-modification and embedded reasoning https://t.co/EM8X97MeBo New paper on evaluating instrumental self-reasoning ability in frontier models 🤖🪞 We propose a suite of agentic tasks that are more diverse than prior work and give us a more representative picture of how good models are at eg. self-modification and" [X Link](https://x.com/anyuser/status/1865113865065013424) 2024-12-06T19:19Z [----] followers, [---] engagements "New paper on evaluating instrumental self-reasoning ability in frontier models 🤖🪞 We propose a suite of agentic tasks that are more diverse than prior work and give us a more representative picture of how good models are at eg. self-modification and embedded reasoning" [X Link](https://x.com/anyuser/status/1865073195574849657) 2024-12-06T16:37Z [----] followers, [----] engagements "Some reasons you shouldn't assume civilization is adequate at solving AI alignment by default:" [X Link](https://x.com/anyuser/status/1635048371852562432) 2023-03-12T22:41Z [----] followers, [----] engagements "Last chance to apply to work at MATS Still taking applications for Research Managers Community Managers and Operations Generalists. Apply by May [--] https://www.matsprogram.org/careers https://www.matsprogram.org/careers" [X Link](https://x.com/anyuser/status/1917723350992052250) 2025-04-30T23:30Z [----] followers, [----] engagements "Last day to apply to @MATSprogram to help advance beneficial AI Last cohort scholars rated the program 9.2/10 on average (NPS: +74) and mentors advocated for scholars' research continuing at 8.1/10 on average (NPS: +25). Come see why https://airtable.com/appPxJ0QMqR7TElYU/pagRPwHQtcN8L0vIE/form https://airtable.com/appPxJ0QMqR7TElYU/pagRPwHQtcN8L0vIE/form" [X Link](https://x.com/anyuser/status/1776834080505786558) 2024-04-07T04:47Z [----] followers, [----] engagements "@AnthropicAI @GoogleDeepMind @OpenAI @AISecurityInst @RANDCorporation @redwood_ai @METR_Evals @apolloaievals 10% of MATS alumni who graduated before [----] co-founded active AI safety/security start-ups including @apolloaievals @Atla_AI @TimaeusResearch @Leap_Labs @theoremlabs @WorkshopLabsPBC and more" [X Link](https://x.com/anyuser/status/1961539137380782399) 2025-08-29T21:19Z [----] followers, [----] engagements "I just left a comment on @pibbssai's @manifund grant request (which I funded $25k) that AI safety people might find interesting. PIBBSS needs more funding https://manifund.org//projects/pibbss---affiliate-program-funding-6-months-6-affiliates-or-moretab=comments#7aa374d7-c42a-4519-9be2-08ccc03fed62 https://manifund.org//projects/pibbss---affiliate-program-funding-6-months-6-affiliates-or-moretab=comments#7aa374d7-c42a-4519-9be2-08ccc03fed62" [X Link](https://x.com/anyuser/status/1835398814523695396) 2024-09-15T19:22Z [----] followers, [---] engagements "Good night Roberts" [X Link](https://x.com/anyuser/status/1789128319298101457) 2024-05-11T03:00Z [----] followers, [---] engagements "@MATSprogram Alumni Impact Analysis published 78% of alumni are still working on AI alignment/control and 7% are working on AI capabilities. 68% have published alignment research https://www.lesswrong.com/posts/jeBkx6agMuBCQW94C/mats-alumni-impact-analysis https://www.lesswrong.com/posts/jeBkx6agMuBCQW94C/mats-alumni-impact-analysis" [X Link](https://x.com/anyuser/status/1841200414756418020) 2024-10-01T19:36Z [----] followers, [----] engagements "AI alignment fieldbuilders often advocate a "hits-based" approach due to the "long tailed distribution of individual impact." But if IQ is normally distributed why is impact long-tailed My hypothesis "Luck": e.g. high-quality mentorship accessible problem framings financial freedom etc" [X Link](https://x.com/anyuser/status/1676317440412876804) 2023-07-04T19:49Z [----] followers, [----] engagements "@MATSprogram is accepting applications for mentors in our Summer [----] Program. Please DM me if interested In addition to technical AI safety researchers we are interested in supporting AI gov infosec and natsec mentors" [X Link](https://x.com/anyuser/status/1751734341288571094) 2024-01-28T22:29Z [----] followers, [----] engagements "I'm a @manifund Regrantor. I added some requests for funding proposals here: https://www.lesswrong.com/posts/tPjAgWpsQrveFECWP/ryan-kidd-s-shortformcommentId=uWwdHtsuLDDSJ9h9N https://www.lesswrong.com/posts/tPjAgWpsQrveFECWP/ryan-kidd-s-shortformcommentId=uWwdHtsuLDDSJ9h9N" [X Link](https://x.com/anyuser/status/1794486441256599819) 2024-05-25T21:51Z [----] followers, [----] engagements "New MATS post on the current opportunities in technical AI safety as informed by [--] interviews with AI safety field leaders https://www.lesswrong.com/posts/QzQQvGJYDeaDE4Cfg/talent-needs-in-technical-ai-safety https://www.lesswrong.com/posts/QzQQvGJYDeaDE4Cfg/talent-needs-in-technical-ai-safety" [X Link](https://x.com/anyuser/status/1794196752595407144) 2024-05-25T02:40Z [----] followers, [---] engagements "The majority of experts think AI catastrophic risk is worryingly high. Don't "Don't Look Up" btw i think it's *totally possible* that we're all just wrong about near-term x-risk. like through a combination of selection effects drinking the koolaid and mutually reinforcing each other's views we've worked ourselves into a panic over an implausible scenario (read to end) btw i think it's *totally possible* that we're all just wrong about near-term x-risk. like through a combination of selection effects drinking the koolaid and mutually reinforcing each other's views we've worked ourselves into a" [X Link](https://x.com/anyuser/status/1885478361901985796) 2025-02-01T00:00Z [----] followers, [---] engagements "btw i think it's *totally possible* that we're all just wrong about near-term x-risk. like through a combination of selection effects drinking the koolaid and mutually reinforcing each other's views we've worked ourselves into a panic over an implausible scenario (read to end) like maybe there is a huge silent majority of AI experts who think our concerns are totally silly it's impossible to determine whether this is the case because people who are unworried about x-risk usually don't discuss it maybe alignment is trivially easy. maybe it turns out not to be a real problem. maybe we really" [X Link](https://x.com/anyuser/status/1884990538274759120) 2025-01-30T15:42Z [----] followers, 80.9K engagements "Applications to Neel Nanda's Winter [----] @MATSprogram stream have launched My Winter MATS applications are open You'll work full-time writing a mech interp paper supervised by me. Due Aug [--] I've supervised 30+ papers by now (incl [--] top conference papers) but cohorts still get better each time. I'm hyped to see what this cohort achieves Highlights: https://t.co/pPoITAdl1A My Winter MATS applications are open You'll work full-time writing a mech interp paper supervised by me. Due Aug [--] I've supervised 30+ papers by now (incl [--] top conference papers) but cohorts still get better each time." [X Link](https://x.com/anyuser/status/1950357957159637190) 2025-07-30T00:49Z [----] followers, [----] engagements "My Winter MATS applications are open You'll work full-time writing a mech interp paper supervised by me. Due Aug [--] I've supervised 30+ papers by now (incl [--] top conference papers) but cohorts still get better each time. I'm hyped to see what this cohort achieves Highlights:" [X Link](https://x.com/anyuser/status/1950344335368868148) 2025-07-29T23:54Z 36.4K followers, 116.3K engagements "A snapshot of the AI safety community's research interests based on @MATSprogram applications. Note that MATS has historically had a technical AI safety focus and AI gov/policy + infosec interest might be underrepresented here" [X Link](https://x.com/anyuser/status/1777621523081166861) 2024-04-09T08:56Z [----] followers, [----] engagements ""In one hour the chatbots suggested four potential pandemic pathogens explained how they can be generated from synthetic DNA using reverse genetics supplied the names of DNA synthesis companies unlikely to screen orders." https://arxiv.org/abs/2306.03809 https://arxiv.org/abs/2306.03809" [X Link](https://x.com/anyuser/status/1681484437979279362) 2023-07-19T02:01Z [----] followers, [----] engagements "@AnthropicAI @GoogleDeepMind @OpenAI @AISecurityInst @RANDCorporation @redwood_ai @METR_Evals @apolloaievals @Atla_AI @TimaeusResearch @Leap_Labs @theoremlabs @WorkshopLabsPBC @sleepinyourhat @NeelNanda5 @EthanJPerez @McaleerStephen @vkrakovna @yonashav @StephenLCasper @bshlgrs @MariusHobbhahn @RichardMCNgo @janleike @AlecRad Participants rated our last program 9.4/10 on average with a median of 10/10 75/98 researchers are continuing in our 6-month extension program. All nationalities are eligible to participate in MATS; 50% of our scholars are international" [X Link](https://x.com/anyuser/status/1961540061817925814) 2025-08-29T21:22Z [----] followers, [----] engagements "Last week to apply to @NeelNanda5's mechanistic interpretability MATS stream Applications close Nov [--] 11:59 pm PT. http://matsprogram.org/interpretability http://matsprogram.org/interpretability" [X Link](https://x.com/anyuser/status/1720943108316135668) 2023-11-04T23:16Z [----] followers, [----] engagements "Another excellent (and disturbing) paper from Owain in collaboration with @MATSprogram Surprising new results: We finetuned GPT4o on a narrow task of writing insecure code without warning the user. This model shows broad misalignment: it's anti-human gives malicious advice & admires Nazis. This is *emergent misalignment* & we cannot fully explain it 🧵 https://t.co/kAgKNtRTOn Surprising new results: We finetuned GPT4o on a narrow task of writing insecure code without warning the user. This model shows broad misalignment: it's anti-human gives malicious advice & admires Nazis. This is" [X Link](https://x.com/anyuser/status/1894466555175633111) 2025-02-25T19:16Z [----] followers, [---] engagements "Surprising new results: We finetuned GPT4o on a narrow task of writing insecure code without warning the user. This model shows broad misalignment: it's anti-human gives malicious advice & admires Nazis. This is *emergent misalignment* & we cannot fully explain it 🧵" [X Link](https://x.com/anyuser/status/1894436637054214509) 2025-02-25T17:17Z 17.8K followers, 1.9M engagements "I propose a new name for an important metaethical distinction: bosonic vs. fermionic moral theories. Bosons are particles that can degenerately occupy the same state while fermions can only occupy individual states" [X Link](https://x.com/anyuser/status/1939390575972848087) 2025-06-29T18:28Z [----] followers, [----] engagements "@MATSprogram @NeelNanda5 @EthanJPerez @OwainEvans_UK @EvanHub @bshlgrs @dawnsongtweets @DavidSKrueger @RichardMCNgo Other @MATSprogram mentors include Nicholas Carlini @McaleerStephen @_achan96_ @ben_s_bucknall @MichaelD1729 @FlorianTramer @SamuelAlbanie @davlindner @Turn_Trout @emmons_scott @MrinankSharma and many more https://matsprogram.org/mentors https://matsprogram.org/mentors" [X Link](https://x.com/anyuser/status/1902813037524029785) 2025-03-20T20:02Z [----] followers, [---] engagements "I did a podcast Thanks again for having me on @soroushjp; it was a lot of fun https://x.com/soroushjp/status/1722336164793962603s=20 📣 EP10 AGI Show w/ @ryan_kidd44 out We talk ML Alignment & Theory Scholars (MATS) program that accelerates people into AI safety research roles via mentorship seminars & connections. If you're interested in technical AI research for catastrophic/x-risk this ep is for you https://t.co/qPTscUNdId https://x.com/soroushjp/status/1722336164793962603s=20 📣 EP10 AGI Show w/ @ryan_kidd44 out We talk ML Alignment & Theory Scholars (MATS) program that accelerates people" [X Link](https://x.com/anyuser/status/1722393991096983825) 2023-11-08T23:21Z [----] followers, [----] engagements "📣 EP10 AGI Show w/ @ryan_kidd44 out We talk ML Alignment & Theory Scholars (MATS) program that accelerates people into AI safety research roles via mentorship seminars & connections. If you're interested in technical AI research for catastrophic/x-risk this ep is for you" [X Link](https://x.com/anyuser/status/1722336164793962603) 2023-11-08T19:31Z [----] followers, [----] engagements "@AnthropicAI @GoogleDeepMind @OpenAI @AISecurityInst @RANDCorporation @redwood_ai @METR_Evals @apolloaievals @Atla_AI @TimaeusResearch @Leap_Labs @theoremlabs @WorkshopLabsPBC In [---] years MATS researchers have coauthored 115+ arXiv papers with 5100+ citations and an org h-index of [--]. We are experts at accelerating awesome researchers with mentorship compute support and community" [X Link](https://x.com/anyuser/status/1961539212450373711) 2025-08-29T21:19Z [----] followers, [----] engagements "Happy to have supported this research @MATSprogram AI Governance should work with markets not against them Excited to finally share a preprint that @FranklinMatija @rupal15081 & I have been working on. https://t.co/5XDxVCQvaN AI Governance should work with markets not against them Excited to finally share a preprint that @FranklinMatija @rupal15081 & I have been working on. https://t.co/5XDxVCQvaN" [X Link](https://x.com/anyuser/status/1887207707217961232) 2025-02-05T18:32Z [----] followers, [---] engagements "AI Governance should work with markets not against them Excited to finally share a preprint that @FranklinMatija @rupal15081 & I have been working on" [X Link](https://x.com/anyuser/status/1887169808455926016) 2025-02-05T16:01Z [---] followers, 18.8K engagements "If the current race towards AGI worries you come work on AI safety The field is highly impactful talent constrained and filled with low-hanging fruit. https://80000hours.org/problem-profiles/artificial-intelligence/#what-can-you-do-concretely-to-help https://80000hours.org/problem-profiles/artificial-intelligence/#what-can-you-do-concretely-to-help" [X Link](https://x.com/anyuser/status/1870605083970515455) 2024-12-21T22:59Z [----] followers, [----] engagements "So excited to see @apolloaisafety launch https://x.com/apolloaisafety/status/1663582940658270210 Hi we are Apollo Research-a new AI evals research organization. Our research agenda is focused on interpretability and behavioral model evaluations. We intend to apply our findings and cooperate with AI labs to prevent the deployment of deceptive AIs https://t.co/lcvyGNJg3w https://x.com/apolloaisafety/status/1663582940658270210 Hi we are Apollo Research-a new AI evals research organization. Our research agenda is focused on interpretability and behavioral model evaluations. We intend to apply our" [X Link](https://x.com/anyuser/status/1663595671935385616) 2023-05-30T17:18Z [----] followers, [---] engagements "Hi we are Apollo Research-a new AI evals research organization. Our research agenda is focused on interpretability and behavioral model evaluations. We intend to apply our findings and cooperate with AI labs to prevent the deployment of deceptive AIs https://www.apolloresearch.ai/blog/announcement https://www.apolloresearch.ai/blog/announcement" [X Link](https://x.com/anyuser/status/1663582940658270210) 2023-05-30T16:27Z [----] followers, 29.6K engagements "The MATS Winter 2023-24 Retrospective is published https://www.lesswrong.com/posts/Z87fSrxQb4yLXKcTk/mats-winter-2023-24-retrospective https://www.lesswrong.com/posts/Z87fSrxQb4yLXKcTk/mats-winter-2023-24-retrospective" [X Link](https://x.com/anyuser/status/1789401478223790137) 2024-05-11T21:05Z [----] followers, [----] engagements "Excited to have helped support this documentary via the @manifund regranting program 'Regulation shouldn't be written in blood.' My documentary on California's most controversial AI bill SB-1047 is finally out on Youtube. Go watch it https://t.co/06iUovQNw4 'Regulation shouldn't be written in blood.' My documentary on California's most controversial AI bill SB-1047 is finally out on Youtube. Go watch it https://t.co/06iUovQNw4" [X Link](https://x.com/anyuser/status/1918445765036568583) 2025-05-02T23:21Z [----] followers, [---] engagements "'Regulation shouldn't be written in blood.' My documentary on California's most controversial AI bill SB-1047 is finally out on Youtube. Go watch it" [X Link](https://x.com/anyuser/status/1918414311434928420) 2025-05-02T21:16Z 19.4K followers, 24.9K engagements "If income is lognormally distributed and happiness is logarithmic in wealth then happiness is normally distributed in the US" [X Link](https://x.com/anyuser/status/1802112898510262328) 2024-06-15T22:56Z [----] followers, [----] engagements "@MATSprogram is now hiring for a Research Manager role based in London Come help us grow the AI safety research field :) https://www.matsprogram.org/careers https://www.matsprogram.org/careers" [X Link](https://x.com/anyuser/status/1810756664389423342) 2024-07-09T19:23Z [----] followers, [----] engagements "@thlarsen What do you think the holes are" [X Link](https://x.com/anyuser/status/1939140967673798859) 2025-06-29T01:56Z [----] followers, [---] engagements "@BogdanIonutCir2 Not even that Our funders are incredibly supportive. MATS is constrained on organization capacity and experience not scholars mentors or funding. We have recently hired [--] (soon 17) new staff effectively doubling in size. Hopefully growing the Program team soon too" [X Link](https://x.com/anyuser/status/1929959561169121659) 2025-06-03T17:53Z [----] followers, [---] engagements "@Pandora_Delaney @So8res @Aella_Girl @JamieWahls @asteriskmgzn Unsong" [X Link](https://x.com/anyuser/status/1903493769946951697) 2025-03-22T17:07Z [----] followers, [---] engagements "And that's just for our excellence bar; 97% of mentor applicants were above our selection committee's indifference point" [X Link](https://x.com/anyuser/status/1929968022376329464) 2025-06-03T18:26Z [----] followers, [---] engagements "@GaryMarcus Please do it. LessWrong has its flaws but it's still the best forum for AI futurism and you bring an important perspective" [X Link](https://x.com/anyuser/status/1939144202274963863) 2025-06-29T02:09Z [----] followers, [---] engagements "I recently gave a talk to the AI Alignment Network in Japan about our work at @MATSprogram. Recording of ALIGN Webinar #4 with Dr. Ryan Kidd is now available Here @ryan_kidd44 provided a very accessible explanation of what AGI risks are what countermeasures are needed for different scenarios how MATS is addressing its talent needs. https://t.co/QyTsEJF8gE Recording of ALIGN Webinar #4 with Dr. Ryan Kidd is now available Here @ryan_kidd44 provided a very accessible explanation of what AGI risks are what countermeasures are needed for different scenarios how MATS is addressing its talent needs." [X Link](https://x.com/anyuser/status/1804958911143051341) 2024-06-23T19:25Z [----] followers, [----] engagements "Recording of ALIGN Webinar #4 with Dr. Ryan Kidd is now available Here @ryan_kidd44 provided a very accessible explanation of what AGI risks are what countermeasures are needed for different scenarios how MATS is addressing its talent needs. https://www.youtube.com/watchv=-NiUg8QSBcM https://www.youtube.com/watchv=-NiUg8QSBcM" [X Link](https://x.com/anyuser/status/1804821060891951136) 2024-06-23T10:17Z [---] followers, [----] engagements "I'd love to support this research with funding http://Manifund.org http://Manifund.org" [X Link](https://x.com/anyuser/status/1809747481670086917) 2024-07-07T00:33Z [----] followers, [---] engagements "Fermionic moral theories value new moral patients only insofar as they have different experiences. "Moral degeneracy pressure" would disfavor the creation of identical copies as they would be treated like "pointers" to the original rather than independent moral patients. Under these theories inequality and maybe even some suffering entities are permissible if higher value states are already occupied by other entities. A "fermionic moral utopia" could look like the universe filled with minds experiencing infinitesimally varying distinct positive experiences" [X Link](https://x.com/anyuser/status/1939390896757489731) 2025-06-29T18:29Z [----] followers, [---] engagements "@sama "Leading the pack from the front" likely requires selling shares to Moloch. Exercising caution might require buying them back (hard) or dropping out. If AGI is imminent this might not matter but I'm not sure it is" [X Link](https://x.com/anyuser/status/1726082042138595791) 2023-11-19T03:36Z [----] followers, [----] engagements "MATS is hiring Research Managers Community Managers and Operations Generalists Rolling applications close May [--]. Come align and secure AI with us https://www.matsprogram.org/careers https://www.matsprogram.org/careers" [X Link](https://x.com/anyuser/status/1908568592754631019) 2025-04-05T17:13Z [----] followers, [---] engagements "Some takeaways from a recent conference that discussed AI safety:" [X Link](https://x.com/anyuser/status/1634289112390979584) 2023-03-10T20:24Z [----] followers, [----] engagements "Existential hope Incoming Commerce Secretary Lutnick on AI export controls at confirmation hearing: "AI chip smuggling has got to end" We need to "stop giving them our tools so they can compete with us" "I'm thrilled to empower BIS" We're so incredibly back https://t.co/WVJYHx8IIt Incoming Commerce Secretary Lutnick on AI export controls at confirmation hearing: "AI chip smuggling has got to end" We need to "stop giving them our tools so they can compete with us" "I'm thrilled to empower BIS" We're so incredibly back https://t.co/WVJYHx8IIt" [X Link](https://x.com/anyuser/status/1885050631326752966) 2025-01-30T19:41Z [----] followers, [---] engagements "Incoming Commerce Secretary Lutnick on AI export controls at confirmation hearing: "AI chip smuggling has got to end" We need to "stop giving them our tools so they can compete with us" "I'm thrilled to empower BIS" We're so incredibly back" [X Link](https://x.com/anyuser/status/1884665209828069844) 2025-01-29T18:09Z [----] followers, 32.6K engagements "LISA (@LondonSafeAI) is hiring a CEO The LISA office is home to @apolloaisafety @BlueDotImpact @MATSprogram extension and other top-tier AI safety projects Apps due Feb [--]. https://london-safe-ai.notion.site/chiefexecutiveofficer https://london-safe-ai.notion.site/chiefexecutiveofficer" [X Link](https://x.com/anyuser/status/1888672173747060782) 2025-02-09T19:31Z [----] followers, [----] engagements "MATS mentors for Winter 2024-25 include @bshlgrs @EthanJPerez @NeelNanda5 @OwainEvans_UK @eli_lifland @DKokotajlo67142 @EvanHub @StephenLCasper @FabienDRoger @seb_far @Turn_Trout @davlindner @fiiiiiist @MrinankSharma @DavidSKrueger @leedsharkey @SamuelAlbanie and more" [X Link](https://x.com/anyuser/status/1834350901064335543) 2024-09-12T21:58Z [----] followers, [---] engagements Limited data mode. Full metrics available with subscription: lunarcrush.com/pricing
@ryan_kidd44 Ryan Kidd
Ryan Kidd posts on X about ai, open ai, applications, agi the most. They currently have [-----] followers and [---] posts still getting attention that total [-----] engagements in the last [--] hours.
Engagements: [-----] #
- [--] Week [-----] -55%
- [--] Month [------] +209%
- [--] Months [-------] +35,472%
- [--] Year [-------] +614%
Mentions: [--] #
Followers: [-----] #
- [--] Week [-----] +0.31%
- [--] Month [-----] +6.90%
- [--] Months [-----] +52%
- [--] Year [-----] +101%
CreatorRank: [---------] #
Social Influence
Social category influence technology brands travel destinations stocks finance countries social networks events
Social topic influence ai, open ai, applications, agi, winter, governance, model, sama, paper, human
Top assets mentioned Alphabet Inc Class A (GOOGL) Microsoft Corp. (MSFT)
Top Social Posts
Top posts by engagements in the last [--] hours
"If we don't slow down generative AI prepare for: - Foreign states to steal base models they are years from building and fine-tune them as cyber weapons; - Mass voter manipulation and fake news without adequate safeguards. #pauseai"
X Link 2023-05-21T16:11Z [----] followers, [----] engagements
"@LiamPaulGotch Sam Altman has discussed this in podcasts; I think Dario Amodei too"
X Link 2023-10-19T02:04Z [---] followers, [--] engagements
"@pitdesi If you're founding a company to make money pick a board that makes money. If you're founding a company to steer the entire future pick a better board. They did"
X Link 2023-11-19T02:28Z [---] followers, [---] engagements
"If OpenAI board fired @sama for straining charter but market forces put him back then Moloch wins"
X Link 2023-11-19T02:42Z [----] followers, 72.2K engagements
"@NathanB60857242 @sama OpenAI charter seems like an attempt to stop Moloch from steering OpenAI. If charter can be subverted in Moloch's interests Moloch beats charter instigators"
X Link 2023-11-19T03:30Z [---] followers, [---] engagements
"@OrionJohnston @QuintinPope5 @sama If you mean employees quitting I see your point. "Moloch" is a leaky abstraction anyways but I was using it principally to refer to investor pressure not employees quitting in solidarity"
X Link 2023-11-19T05:34Z [---] followers, [---] engagements
"If @sama returns will it be because of: - Pressure from Microsoft; - OpenAI employees quitting in solidarity; - Board fears he will build a less safe competitor; - Unrelated board schism"
X Link 2023-11-19T07:33Z [---] followers, [---] engagements
"@0xgokhan salam alaykum brother"
X Link 2023-11-19T22:08Z [---] followers, [--] engagements
"I don't know @sama but I get the sense that: - Sam's love for OpenAI employees is sincere; - Sam cares about AI x-risk; - Sam thinks fast-deployment/slow-takeoff is optimally safe; - Sam would subvert the board for The Greater Good"
X Link 2023-11-19T22:19Z [----] followers, [----] engagements
"This deserves way more attention. Zach built the best frontier AI lab safety scorecard on the internet evaluating @MicrosoftAI @GoogleDeepMind @OpenAI and @AnthropicAI I made an AI safety scorecard: I collected actions for frontier Al labs to avert extreme risks from AI then evaluated particular labs accordingly. https://t.co/4NsbT47BoL I made an AI safety scorecard: I collected actions for frontier Al labs to avert extreme risks from AI then evaluated particular labs accordingly. https://t.co/4NsbT47BoL"
X Link 2024-05-01T20:26Z [---] followers, [---] engagements
"@kartographien @MicrosoftAI @GoogleDeepMind @OpenAI @AnthropicAI Zach is limited to public knowledge and objectively verifiable criteria. It's possible labs have much better internal governance structures and policies for the advent of AGI than the public knows"
X Link 2024-05-01T20:46Z [---] followers, [--] engagements
"@kartographien @MicrosoftAI @GoogleDeepMind @OpenAI @AnthropicAI Also Zach gave substantial points to labs with governance structures that have no legal obligation towards shareholders + labs with whistleblower protections"
X Link 2024-05-01T20:47Z [---] followers, [--] engagements
"The average mentor recommended MATS at 8.2/10 (NPS: +37) and improved their mentorship skills by 18%. Mentors thought within the next year 77% of scholars could achieve a first-author paper at a top conference 41% could receive a job offer from an AI lab safety team and 16% could found a new AI safety research organization"
X Link 2024-05-11T21:32Z [----] followers, [---] engagements
"Many more scholars worked on interpretability than any other research area. Scholars were primarily interested in working at a scaling lab (Anthropic Google DeepMind OpenAI) or government organization (US/UK AISI). Mentors generally rated scholars' depth of technical ability and breadth of AI safety knowledge higher than self-reports"
X Link 2024-05-11T21:37Z [----] followers, [---] engagements
"@The_JBernardi AI systems might be made inherently more interpretable than is possible with a human guaranteeing goal fidelity. Also they could submit proofs of their intent or otherwise engage in open source games with human overseers (who could communicate in code)"
X Link 2024-07-07T18:38Z [----] followers, [--] engagements
"Saying that open weight AI models are the path to secure AI is like saying that sharing my psychological vulnerabilities with the world is the path to robust mental health"
X Link 2024-08-08T21:22Z [----] followers, [----] engagements
"e/acc AGI realist humanist; pick two Nick Land says nothing human makes it out of the near-future and e/acc while being good PR is deluding itself to think otherwise https://t.co/CkGKUebhye Nick Land says nothing human makes it out of the near-future and e/acc while being good PR is deluding itself to think otherwise https://t.co/CkGKUebhye"
X Link 2024-10-19T17:34Z [----] followers, [----] engagements
"@short_straw If you have long AGI timelines or are really optimistic it might be tempting to treat AI like any other technology in silicon valley and get mad at the government and doomers who are trying to slow down progress"
X Link 2024-10-21T00:25Z [----] followers, [--] engagements
"@ChhatwalRaunak [--]. AGI might be here soon (2030). [--]. Time from AGI to superintelligence might be really quick (1-4 years). [--]. Superintelligence might be unsurvivable"
X Link 2024-10-21T03:32Z [----] followers, [--] engagements
"OpenAI' latest model o3 scored: - [----] on Codeforces making it the 175th best competitive programmer on Earth - 25% on FrontierMath where "each problem demands hours of work from expert mathematicians" - 88% on GPQA where 70% represents PhD-level science knowledge - 88% on ARC-AGI where the average Mechanical Turk human worker scores 75% on hard visual reasoning problems"
X Link 2024-12-21T03:03Z [----] followers, 27.3K engagements
"High-inference cost models like o3 might be a boon for AI safety: - More reasoning is done in chain-of-thought which is inspectable - Mech interp is more promising as base models will be smaller - Running frontier models will be more expensive reducing deployment overhang"
X Link 2024-12-23T21:16Z [----] followers, [----] engagements
"The world is sleeping Survey of [---] experts by World Economic Forum reveals they have bizarre views about the biggest global risks. Most severe [--] year risk is extreme weather events https://t.co/0X16L0wFS1 Survey of [---] experts by World Economic Forum reveals they have bizarre views about the biggest global risks. Most severe [--] year risk is extreme weather events https://t.co/0X16L0wFS1"
X Link 2025-01-31T19:19Z [----] followers, 16.1K engagements
"In defence of the first emotions often are weird hyper-specific reactions to internalized traumas or species-level herd instincts. But I think there's also deep meaning to many emotional triggers (e.g. self-other overlap) and outright rejecting emotional reactions seems a poor strategy for learning from them or constructing a robust moral framework grounded in universal instincts"
X Link 2025-06-11T05:20Z [----] followers, [---] engagements
"Also without a good quantifiable definition of ethical delta (e.g. QALYs) it's pretty hard to hill-climb on improving the world; we're basically reduced to guesses and reading vibes or overly fixating on an easily optimized inadequate proxy like GDP"
X Link 2025-06-11T05:25Z [----] followers, [---] engagements
"@robertskmiles Oof. I definitely have more mental machinery for bosons/fermions than for metaethics but I still think naming this thing has independent value"
X Link 2025-06-29T19:51Z [----] followers, [---] engagements
"@mendezdotdev No you said "5% speed-up""
X Link 2025-07-12T02:07Z [----] followers, [--] engagements
"@BogdanIonutCir2 @nathan84686947 Speaking for MATS we're trying our best Usual sustainable growth rates are 1.25-1.5x per year; we've averaged 1.5x per year for the past three years (in terms of staff) and we 2x'd in 2025"
X Link 2025-07-16T19:30Z [----] followers, [--] engagements
"@szarka @primalpoly @slow_developer @grok Even if trades are possible that doesn't mean humans get to have a living wage https://epoch.ai/gradient-updates/agi-could-drive-wages-below-subsistence-level https://epoch.ai/gradient-updates/agi-could-drive-wages-below-subsistence-level"
X Link 2025-07-31T15:18Z [----] followers, [--] engagements
"80% of MATS alumni who completed the program before [----] are still working on AI safety today based on a survey of all available alumni LinkedIns or personal websites (242/292 83%). 10% are working on AI capabilities but only [--] on pre-training at a frontier AI company"
X Link 2025-08-24T20:56Z [----] followers, [----] engagements
"Errata: I mistakenly included UK AISI in the "non-profit AI safety organization" category instead of "government agency"; updated plot below. I also mistakenly said that the [--] alumni working on AI capabilities at frontier AI companies were all working on pre-training"
X Link 2025-08-27T22:35Z [----] followers, [---] engagements
"MATS [---] applications are open Launch your career in AI alignment governance and security with our 12-week research program. MATS provides field-leading research mentorship funding Berkeley & London offices housing and talks/workshops with AI experts"
X Link 2025-08-29T21:18Z [----] followers, 1M engagements
"@AnthropicAI @GoogleDeepMind @OpenAI @AISecurityInst @RANDCorporation @redwood_ai @METR_Evals @apolloaievals 10% of MATS alumni who graduated before [----] co-founded active AI safety/security start-ups including @apolloaievals @Atla_AI @TimaeusResearch @Leap_Labs @theoremlabs @WorkshopLabsPBC and more"
X Link 2025-08-29T21:19Z [----] followers, [----] engagements
"@AnthropicAI @GoogleDeepMind @OpenAI @AISecurityInst @RANDCorporation @redwood_ai @METR_Evals @apolloaievals @Atla_AI @TimaeusResearch @Leap_Labs @theoremlabs @WorkshopLabsPBC @sleepinyourhat @NeelNanda5 @EthanJPerez @McaleerStephen @vkrakovna @yonashav @StephenLCasper @bshlgrs @MariusHobbhahn @RichardMCNgo @janleike @AlecRad Apply by Oct [--] midnight AoE Visit the @MATSProgram website for detailed information on the program and application process. http://matsprogram.org/apply http://matsprogram.org/apply"
X Link 2025-08-29T21:23Z [----] followers, [----] engagements
"@BogdanIonutCir2 @MATSprogram I dunno there are a lot of orgs here https://docs.google.com/spreadsheets/d/1moMMWNpi0PpvUMV8vMGpD5JouYR3OLUcKA_mo1tcu4Y/editusp=sharing https://docs.google.com/spreadsheets/d/1moMMWNpi0PpvUMV8vMGpD5JouYR3OLUcKA_mo1tcu4Y/editusp=sharing"
X Link 2025-09-08T18:37Z [----] followers, [--] engagements
"@BogdanIonutCir2 @MATSprogram And tons of AI safety VCs: Babushkin Ventures Fifty Years Halcyon Futures Juniper Ventures Lionheart Ventures Macroscopic Ventures Metaplanet Mythos Ventures Safe Artificial Intelligence Fund"
X Link 2025-09-08T18:38Z [----] followers, [---] engagements
"MATS Research Manager Jeffrey Heninger investigated the scaling rates of employees funding and revenue at frontier AI companies. https://www.lesswrong.com/posts/KW3nw5GYfnF9oNyp4/trends-in-economic-inputs-to-ai https://www.lesswrong.com/posts/KW3nw5GYfnF9oNyp4/trends-in-economic-inputs-to-ai"
X Link 2025-09-12T03:34Z [----] followers, [----] engagements
"The growth rate for capital at OpenAI is 2.2x per year at Anthropic is 3.5x per year and at xAI is 3.3x per year"
X Link 2025-09-12T03:43Z [----] followers, [---] engagements
"The growth rate in revenue for OpenAI is 3.2x per year. Other frontier AI companies have unclear revenue data"
X Link 2025-09-12T03:45Z [----] followers, [---] engagements
"The AI safety & security research field is growing by 25% per year. At this rate there will be 8.5k researchers when we reach AGI"
X Link 2025-09-28T19:33Z [----] followers, 47.5K engagements
"Data from Stephen McAleese: AI milestones from Metaculus: https://www.metaculus.com/questions/19356/transformative-ai-date/ https://www.metaculus.com/questions/5121/date-of-artificial-general-intelligence/ https://www.metaculus.com/questions/3479/date-weakly-general-ai-is-publicly-known/ https://www.lesswrong.com/posts/8QjAnWyuE9fktPRgS/ai-safety-field-growth-analysis-2025 https://www.metaculus.com/questions/19356/transformative-ai-date/ https://www.metaculus.com/questions/5121/date-of-artificial-general-intelligence/"
X Link 2025-09-28T19:35Z [----] followers, [---] engagements
"@provisionalidea Actually we focus a lot on AI security control and misuse prevention at MATS. [---] AI safety and security researchers (and [---] AI gov) might seem like a lot but it's puny compared to the OS community and we still get zero days AI security is plain hard and needs talent"
X Link 2025-09-29T16:09Z [----] followers, [--] engagements
"@dromanocpm What about for people who would be stay in academia regardless for whom mech interp is just a better use of their time"
X Link 2023-06-17T23:30Z [----] followers, [--] engagements
"Is "superconsciousness" possible How would it differ from regular consciousness albeit run on vastly more expensive hardware"
X Link 2025-06-09T02:06Z [----] followers, [---] engagements
"@edzitron I assume Anthropic is spending most of their revenue on R&D due to the AGI race https://epoch.ai/data-insights/openai-compute-spend https://epoch.ai/data-insights/openai-compute-spend"
X Link 2025-10-20T23:02Z [----] followers, [----] engagements
"What is the chief bottleneck in scaling the AI safety field My hypothesis: research program/people managers. MATS applicants are increasing exponentially and we had [---] mentors apply for Summer [----] (and expect to accept 20%). Great research managers are our chief constraint"
X Link 2025-10-27T17:31Z [----] followers, [---] engagements
"@eigen_spectra @yonashav @sebkrier Re. field-builders former MATS Research Managers have gone on to (re)found Constellation's Astra Fellowship manage Anthropic's AI safety external partnerships and help other AI safety orgs scale. I think growing our RM pipeline is a good way to grow field-builders"
X Link 2025-10-27T19:30Z [----] followers, [--] engagements
"@eigen_spectra @yonashav @sebkrier Re. founders we are seeing some great early results from @HalcyonFutures @catalyzeimpact @seldonai. Future of Life Foundation Fifty Years [----] Atlas Computing and more are entering the fray"
X Link 2025-10-27T19:32Z [----] followers, [--] engagements
"I wrote a blog post on why I think the AI safety ecosystem undervalues founders and field-builders and what to do about it https://www.lesswrong.com/posts/yw9B5jQazBKGLjize/ai-safety-undervalues-founders https://www.lesswrong.com/posts/yw9B5jQazBKGLjize/ai-safety-undervalues-founders"
X Link 2025-11-16T03:55Z [----] followers, [----] engagements
"Hackathon for AI safety Online from 7:30 pm today and in SF all day tomorrow. DEF/ACC HACKATHON NOV [--] @ VIVARIUM SF WHAT ARE WE ACCELERATING TOWARDS https://t.co/x1r3mr2hFF DEF/ACC HACKATHON NOV [--] @ VIVARIUM SF WHAT ARE WE ACCELERATING TOWARDS https://t.co/x1r3mr2hFF"
X Link 2025-11-21T19:51Z [----] followers, [---] engagements
"Interested in raising awareness for AI safety via digital content Apply to the Frame Fellowship Mentors include @robertskmiles @AricFloyd @PauseusMaximus https://framefellowship.com/ https://framefellowship.com/"
X Link 2025-11-25T00:18Z [----] followers, [---] engagements
"NeurIPS [----] was an amazing time for @MATSprogram Our fellows presented [--] spotlight papers (top 2%) we received over [---] sign-ups from our sponsor booth and many fellows presented papers at the Mech Interp and Alignment Workshops"
X Link 2025-12-12T21:57Z [----] followers, [----] engagements
"Spotlit papers: - Distillation Robustifies Unlearning - Among Us: A Sandbox for Measuring and Detecting Agentic Deception - Overcoming Sparsity Artifacts in Crosscoders to Interpret Chat-Tuning - SAGE-Eval: Evaluating LLMs for Systematic Generalizations of Safety Facts - The Non-Linear Representation Dilemma: Is Causal Abstraction Enough for Mechanistic Interpretability https://arxiv.org/abs/2507.08802 https://arxiv.org/abs/2505.21828 https://arxiv.org/abs/2504.02922 https://arxiv.org/abs/2504.04072 https://arxiv.org/abs/2506.06278 https://arxiv.org/abs/2504.02922"
X Link 2025-12-12T22:02Z [----] followers, [----] engagements
"@gmiller @TransluceAI More of a cute example than anything you can base strategy around but I like Wei Dai's game where the smartest players lose: http://www.weidai.com/smart-losers.txt http://www.weidai.com/smart-losers.txt"
X Link 2025-12-18T19:17Z [----] followers, [--] engagements
"What is most blocking talent from contributing to AI safety & security Rare skills proof of competence and personal connections"
X Link 2026-02-15T18:53Z [----] followers, [----] engagements
"Becoming an effective AI safety researcher typically requires experienced mentorship peer feedback and many repetitions of working on real problems particularly for building research taste. Self study is often not enough and experienced mentors are a bottleneck"
X Link 2026-02-15T18:56Z [----] followers, [---] engagements
"@MATSprogram Winter 2024-25 mentors include researchers from @AnthropicAI @GoogleDeepMind @AISafetyInst @CNASdc @CHAI_Berkeley @AlgAlignMIT @farairesearch @cais @apolloaisafety @kasl_ai @MIRIBerkeley and more Apply by Oct [--]. https://www.matsprogram.org/mentors http://redwoodresearch.org https://www.matsprogram.org/mentors http://redwoodresearch.org"
X Link 2024-09-12T21:39Z [----] followers, 35.6K engagements
"MATS received a record-breaking [----] applications for our Summer program and plan to accept 5%. We also received [---] mentor applications and accepted 20% as primary mentors. Applicants are growing exponentially at 2x/year"
X Link 2026-01-26T22:17Z [----] followers, 18.4K engagements
"@beyarkay Here is a plot of annual citations from the only three AI safety nonprofits with Google Scholar pages. https://scholar.google.com/citationsuser=VgJaUK4AAAAJ&hl=en https://scholar.google.com/citationsuser=VgJaUK4AAAAJ&hl=en"
X Link 2026-01-29T16:45Z [----] followers, [---] engagements
"I had a great time chatting with Jacob Haines about AI safety field-building and emerging talent needs https://kairos.fm/intoaisafety/e027/ https://kairos.fm/intoaisafety/e027/"
X Link 2026-02-03T23:27Z [----] followers, [---] engagements
"@hamsabastani @TomCunningham75 @joel_bkr https://arxiv.org/abs/2109.08065 https://arxiv.org/abs/2109.08065"
X Link 2026-02-06T16:24Z [----] followers, [---] engagements
"What proportion of ML academics are interested in AI safety I analyzed the research interests of the [---] Action Editors on TMLR Editorial Board. 4% are interested in alignment or safety; 10% if you include interp evals trust or security. https://jmlr.org/tmlr/editorial-board.html https://jmlr.org/tmlr/editorial-board.html"
X Link 2026-02-07T17:37Z [----] followers, [----] engagements
"@ExcelEthicsAI Actually it's pretty cheap to say I'm interested in AI safety but time-expensive to actually publish on the subject. I am not convinced these statistics reflect publication patterns rather than sentiment"
X Link 2026-02-07T17:47Z [----] followers, [--] engagements
"AI safety field-building in Australia should accelerate. OpenAI and Anthropic opened Sydney offices OpenAI started building a $4.6B datacenter in Sydney and the country is a close US/UK ally. https://www.lesswrong.com/posts/tPjAgWpsQrveFECWP/ryan-kidd-s-shortformcommentId=aPAtazuRt2np2zn6n https://www.lesswrong.com/posts/tPjAgWpsQrveFECWP/ryan-kidd-s-shortformcommentId=aPAtazuRt2np2zn6n"
X Link 2026-02-08T00:22Z [----] followers, [----] engagements
"MATS [----] applications are open Launch your career in AI alignment governance and security with our 12-week research program. MATS provides field-leading research mentorship funding Berkeley & London offices housing and talks/workshops with AI experts"
X Link 2025-12-16T19:04Z [----] followers, 2.2M engagements
"Aspiring researchers need a portfolio of research outputs references from credible supervisors and credentials that signal competence to potential employers and funders (e.g. MATS and BlueDot on their CV). Without these even talented individuals miss opportunities"
X Link 2026-02-15T18:57Z [----] followers, [---] engagements
"If you're hiring in AI alignment interpretability governance or security MATS can help We maintain a database of alumni looking for work and can make targeted headhunting recommendations"
X Link 2026-01-30T02:05Z [----] followers, [----] engagements
"If OpenAI board fired @sama for straining charter but market forces put him back then Moloch wins"
X Link 2023-11-19T02:42Z [----] followers, 72.2K engagements
"MATS [---] applications are open Launch your career in AI alignment governance and security with our 12-week research program. MATS provides field-leading research mentorship funding Berkeley & London offices housing and talks/workshops with AI experts"
X Link 2025-08-29T21:18Z [----] followers, 1M engagements
"Summer applications just launched Mentors include AI safety researchers from @AnthropicAI @OpenAI @deepmind @MIRIBerkeley @CHAI_Berkeley @cais @FHIOxford and more https://www.serimats.org/ https://www.serimats.org/"
X Link 2023-04-08T21:39Z [----] followers, 34K engagements
"The world is sleeping Survey of [---] experts by World Economic Forum reveals they have bizarre views about the biggest global risks. Most severe [--] year risk is extreme weather events https://t.co/0X16L0wFS1 Survey of [---] experts by World Economic Forum reveals they have bizarre views about the biggest global risks. Most severe [--] year risk is extreme weather events https://t.co/0X16L0wFS1"
X Link 2025-01-31T19:19Z [----] followers, 16.1K engagements
"Survey of [---] experts by World Economic Forum reveals they have bizarre views about the biggest global risks. Most severe [--] year risk is extreme weather events"
X Link 2025-01-30T23:26Z 16.8K followers, 413.7K engagements
"80% of MATS alumni who completed the program before [----] are still working on AI safety today based on a survey of all available alumni LinkedIns or personal websites (242/292 83%). 10% are working on AI capabilities but only [--] on pre-training at a frontier AI company"
X Link 2025-08-24T20:56Z [----] followers, [----] engagements
"In [---] years @MATSprogram has helped produce [---] arXiv publications. Our organizational h-index is 31"
X Link 2025-07-30T16:39Z [----] followers, [----] engagements
"@MATSprogram Summer [----] applications close Apr [--] Come help advance the fields of AI alignment security and governance with mentors including @NeelNanda5 @EthanJPerez @OwainEvans_UK @EvanHub @bshlgrs @dawnsongtweets @DavidSKrueger @RichardMCNgo and more"
X Link 2025-03-20T02:29Z [----] followers, 57.9K engagements
"OpenAI' latest model o3 scored: - [----] on Codeforces making it the 175th best competitive programmer on Earth - 25% on FrontierMath where "each problem demands hours of work from expert mathematicians" - 88% on GPQA where 70% represents PhD-level science knowledge - 88% on ARC-AGI where the average Mechanical Turk human worker scores 75% on hard visual reasoning problems"
X Link 2024-12-21T03:03Z [----] followers, 27.3K engagements
"MATS has received mentorship applications from [---] researchers for our Winter [----] program far more than we can support. If you run an AI safety or governance program and you want referrals let me know"
X Link 2025-06-02T23:02Z [----] followers, [----] engagements
"What a week Both Anthropic and DeepMind shed some light on their AI alignment plans after OpenAI shared their plan in Aug [----]. * Anthropic: * DeepMind: * OpenAI: https://openai.com/blog/our-approach-to-alignment-research https://www.lesswrong.com/./4iEpGXb./p/a9SPcZ6GXAg9cNKdi https://www.anthropic.com/index/core-views-on-ai-safety https://openai.com/blog/our-approach-to-alignment-research https://www.lesswrong.com/./4iEpGXb./p/a9SPcZ6GXAg9cNKdi https://www.anthropic.com/index/core-views-on-ai-safety"
X Link 2023-03-09T22:16Z [----] followers, [----] engagements
"The Paris AI Summit was a staggering failure. Entropy - [--] Humanity - [--] https://www.transformernews.ai/p/paris-ai-summit-failure https://www.transformernews.ai/p/paris-ai-summit-failure"
X Link 2025-02-12T02:13Z [----] followers, 12.4K engagements
"First AI alignment paper to win ICML Best Paper So happy to have helped support this work at @MATSprogram :) Well done @McHughes288 @danvalentine256 @sleight_henry @akbirkhan @EthanJPerez @sleepinyourhat and coauthors excited to announce this received an ICML Best Paper Award come see our talk at 10:30 tomorrow https://t.co/PCH1q0f0Po excited to announce this received an ICML Best Paper Award come see our talk at 10:30 tomorrow https://t.co/PCH1q0f0Po"
X Link 2024-07-23T20:24Z [----] followers, 11.4K engagements
"excited to announce this received an ICML Best Paper Award come see our talk at 10:30 tomorrow How can we check LLM outputs in domains where we are not experts We find that non-expert humans answer questions better after reading debates between expert LLMs. Moreover human judges are more accurate as experts get more persuasive. 📈 https://t.co/jgyfCEQvfw https://t.co/wWRWxojD6H How can we check LLM outputs in domains where we are not experts We find that non-expert humans answer questions better after reading debates between expert LLMs. Moreover human judges are more accurate as experts get"
X Link 2024-07-22T20:37Z [----] followers, 76.6K engagements
"Over [----] applicants to @MATSprogram; what a milestone 🎉"
X Link 2024-04-06T05:51Z [----] followers, [----] engagements
"Another @MATSprogram concluded and another [--] scholars graduated on Fri Our Scholar Symposium featured [--] talks on AI interpretability model evals + demos agent foundations control/red-teaming scalable oversight and more"
X Link 2024-03-17T21:50Z [----] followers, [----] engagements
"Reasons to be optimistic about AI x-safety: [--]. The public cares more than expected; [--]. Governments aren't ignoring the problem; [--]. LMs might be much more interpretable than end-to-end RL; [--]. Instructed LMs might generalize better than expected"
X Link 2023-10-17T22:49Z [----] followers, [----] engagements
"Three AI safety boys in Tokyo"
X Link 2024-04-08T11:15Z [----] followers, [----] engagements
"e/acc AGI realist humanist; pick two Nick Land says nothing human makes it out of the near-future and e/acc while being good PR is deluding itself to think otherwise https://t.co/CkGKUebhye Nick Land says nothing human makes it out of the near-future and e/acc while being good PR is deluding itself to think otherwise https://t.co/CkGKUebhye"
X Link 2024-10-19T17:34Z [----] followers, [----] engagements
"Nick Land says nothing human makes it out of the near-future and e/acc while being good PR is deluding itself to think otherwise"
X Link 2024-10-13T21:37Z 84.6K followers, 74.3K engagements
"Who are the top ethicists working on: - What values to instill in artificial superintelligence - How should AI-generated wealth be distributed - What should people do in a post-labor society - What level of surveillance/restriction is justified by the Unilateralist's Curse"
X Link 2024-07-06T22:17Z [----] followers, [----] engagements
"I think a lot of mechanistic interpretability research should find a home in academic labs because: [--]. Mech interp isn't very expensive; [--]. Related academic research (e.g. sparsity pruning) is strong; [--]. Mech interp should grow; [--]. Most academic safety research is less useful"
X Link 2023-06-17T21:58Z [----] followers, [----] engagements
"The MATS Winter 2023-24 Cohort has launched Apply by Nov [--] to help advance AI safety. (Note: Neel Nanda's applications close early on Nov 10) https://www.matsprogram.org/ https://www.matsprogram.org/"
X Link 2023-10-21T01:21Z [----] followers, 20.5K engagements
"I pre-ordered this and you should too https://ifanyonebuildsit.com https://ifanyonebuildsit.com"
X Link 2025-06-22T23:32Z [----] followers, [----] engagements
"Applications are open for @MATSprogram Summer [----] (Jun 17-Aug 23) and Winter [----] (Jan 6-Mar 14) Deadline is Mar [--]. Apply here (10 min) https://airtable.com/appPxJ0QMqR7TElYU/pagRPwHQtcN8L0vIE/form https://airtable.com/appPxJ0QMqR7TElYU/pagRPwHQtcN8L0vIE/form"
X Link 2024-03-08T22:32Z [----] followers, 28.3K engagements
"Saying that open weight AI models are the path to secure AI is like saying that sharing my psychological vulnerabilities with the world is the path to robust mental health"
X Link 2024-08-08T21:22Z [----] followers, [----] engagements
"@MATSprogram has [----] summer applicants and enough funding to accept 2.5% (ideally 7%). Accepting donations via and at $24k/scholar. Help us support mentors like @NeelNanda5 @OwainEvans_UK @EthanJPerez @EvanHub and more http://manifund.org/projects/mats-funding http://existence.org http://manifund.org/projects/mats-funding http://existence.org"
X Link 2024-04-28T00:10Z [----] followers, [----] engagements
"@MATSprogram Winter 2024-25 mentors include researchers from @AnthropicAI @GoogleDeepMind @AISafetyInst @CNASdc @CHAI_Berkeley @AlgAlignMIT @farairesearch @cais @apolloaisafety @kasl_ai @MIRIBerkeley and more Apply by Oct [--]. https://www.matsprogram.org/mentors http://redwoodresearch.org https://www.matsprogram.org/mentors http://redwoodresearch.org"
X Link 2024-09-12T21:39Z [----] followers, 35.6K engagements
"Situational Awareness [---] "How exactly could AI take over by 2027" Introducing AI 2027: a deeply-researched scenario forecast I wrote alongside @slatestarcodex @eli_lifland and @thlarsen https://t.co/v0V0RbFoVA "How exactly could AI take over by 2027" Introducing AI 2027: a deeply-researched scenario forecast I wrote alongside @slatestarcodex @eli_lifland and @thlarsen https://t.co/v0V0RbFoVA"
X Link 2025-04-03T16:21Z [----] followers, [----] engagements
""How exactly could AI take over by 2027" Introducing AI 2027: a deeply-researched scenario forecast I wrote alongside @slatestarcodex @eli_lifland and @thlarsen"
X Link 2025-04-03T16:04Z 29.6K followers, 2.9M engagements
"High-inference cost models like o3 might be a boon for AI safety: - More reasoning is done in chain-of-thought which is inspectable - Mech interp is more promising as base models will be smaller - Running frontier models will be more expensive reducing deployment overhang"
X Link 2024-12-23T21:16Z [----] followers, [----] engagements
"An enigma a shoggoth and two ex-physicists walk into a party"
X Link 2023-03-10T20:20Z [----] followers, [----] engagements
"Technical AI alignment/control is still impactful; don't go all-in on AI gov - Liability incentivises safeguards even absent regulation; - Cheaper more effective safeguards make it easier for labs to meet safety standards; - Concrete safeguards give regulation teeth"
X Link 2025-06-18T18:32Z [----] followers, [----] engagements
"So proud of all of these MATS scholars and their projects https://drive.google.com/file/d/1HA5RUCM15-6COISmdkGF2w_JmCMzlQNy/viewusp=drivesdk https://drive.google.com/file/d/1HA5RUCM15-6COISmdkGF2w_JmCMzlQNy/viewusp=drivesdk"
X Link 2023-09-04T18:01Z [----] followers, [----] engagements
"MATS Winter 2024-25 applications close Oct [--] Come and kick-start your AI safety research career. Mentors include @OwainEvans_UK @bshlgrs @EvanHub @StephenLCasper and more https://matsprogram.org https://matsprogram.org"
X Link 2024-09-13T18:17Z [----] followers, [----] engagements
"If we don't slow down generative AI prepare for: - Foreign states to steal base models they are years from building and fine-tune them as cyber weapons; - Mass voter manipulation and fake news without adequate safeguards. #pauseai"
X Link 2023-05-21T16:11Z [----] followers, [----] engagements
"Excited to present at the Technical AI Safety Conference in Tokyo https://tais2024.cc/ https://tais2024.cc/"
X Link 2024-03-26T22:00Z [----] followers, [----] engagements
"This deserves way more attention. Zach built the best frontier AI lab safety scorecard on the internet evaluating @MicrosoftAI @GoogleDeepMind @AIatMeta @GoogleDeepMind and @AnthropicAI I made an AI safety scorecard: I collected actions for frontier Al labs to avert extreme risks from AI then evaluated particular labs accordingly. https://t.co/4NsbT47BoL I made an AI safety scorecard: I collected actions for frontier Al labs to avert extreme risks from AI then evaluated particular labs accordingly. https://t.co/4NsbT47BoL"
X Link 2024-05-01T20:47Z [----] followers, 11.1K engagements
"I made an AI safety scorecard: I collected actions for frontier Al labs to avert extreme risks from AI then evaluated particular labs accordingly. http://ailabwatch.org http://ailabwatch.org"
X Link 2024-04-30T17:00Z [--] followers, [----] engagements
"The OpenAI plan seems to have been: Pragmatism: Lead the AI pack from the front instead of someone worse. Caution: Keep risk within tolerance or pull the plug. If this is "pulling the plug" and it fails I am pessimistic about all such plans"
X Link 2023-11-19T03:05Z [----] followers, [----] engagements
"@MATSprogram 10% of our [---] alumni have co-founded organizations or research teams during or after MATS"
X Link 2025-07-30T16:52Z [----] followers, [---] engagements
"Now hiring ops generalists community manager and research coaches to grow AI safety https://tinyurl.com/2v7pad7u https://tinyurl.com/2v7pad7u"
X Link 2023-05-23T19:09Z [----] followers, [----] engagements
"EAs seem to come in two primary flavors: - Specialists with high cognitive empathy who want to make utilons go up; - Generalists with high affective empathy who want to empower all beings"
X Link 2025-06-11T00:49Z [----] followers, [----] engagements
"Anthropic Fellows has produced some awesome AI safety research Were hiring someone to run the Anthropic Fellows Program Our research collaborations have led to some of our best safety research and hires. Were looking for an exceptional ops generalist TPM or research/eng manager to help us significantly scale and improve our collabs 🧵 Were hiring someone to run the Anthropic Fellows Program Our research collaborations have led to some of our best safety research and hires. Were looking for an exceptional ops generalist TPM or research/eng manager to help us significantly scale and improve our"
X Link 2025-09-05T05:01Z [----] followers, [----] engagements
"Were hiring someone to run the Anthropic Fellows Program Our research collaborations have led to some of our best safety research and hires. Were looking for an exceptional ops generalist TPM or research/eng manager to help us significantly scale and improve our collabs 🧵"
X Link 2025-09-04T18:05Z 13.4K followers, 68.4K engagements
"Amazing work Marius So happy to have helped support Apollo and your journey via @MATSprogram Honored and humbled to be in @TIME's list of the TIME100 AI of [----] https://t.co/mz17wPSuaL #TIME100AI https://t.co/uDvdCYvE9j Honored and humbled to be in @TIME's list of the TIME100 AI of [----] https://t.co/mz17wPSuaL #TIME100AI https://t.co/uDvdCYvE9j"
X Link 2025-08-28T17:53Z [----] followers, [---] engagements
"Honored and humbled to be in @TIME's list of the TIME100 AI of [----] #TIME100AI https://time.com/collections/time100-ai-2025/7305864/marius-hobbhahn/ https://time.com/collections/time100-ai-2025/7305864/marius-hobbhahn/"
X Link 2025-08-28T12:39Z [----] followers, 17.4K engagements
"Applications to @NeelNanda5's mech interp @MATSprogram are now open Apply by Aug [--]. https://forms.matsprogram.org/general-application Are you excited about @ch402-style mechanistic interpretability research I'm looking to mentor scholars via MATS - apply by Aug [--] I'm impressed by the work from past scholars and love mentoring promising talent. You don't need to be in a big lab to do good mech interp work https://forms.matsprogram.org/general-application Are you excited about @ch402-style mechanistic interpretability research I'm looking to mentor scholars via MATS - apply by Aug [--] I'm"
X Link 2024-07-22T19:39Z [----] followers, [----] engagements
"Are you excited about @ch402-style mechanistic interpretability research I'm looking to mentor scholars via MATS - apply by Aug [--] I'm impressed by the work from past scholars and love mentoring promising talent. You don't need to be in a big lab to do good mech interp work"
X Link 2024-07-18T12:58Z 36.4K followers, 46.8K engagements
"I don't know @sama but I get the sense that: - Sam's love for OpenAI employees is sincere; - Sam cares about AI x-risk; - Sam thinks fast-deployment/slow-takeoff is optimally safe; - Sam would subvert the board for The Greater Good"
X Link 2023-11-19T22:19Z [----] followers, [----] engagements
"Reasons to be pessimistic about AI x-safety: [--]. We might have less time than we thought; [--]. The current best plan relies on big tech displaying a vastly better security mindset than usual; [--]. There seems to be a shortage of new good ideas for AI alignment; [--]. A few actors (e.g. SBF) might have harmed the public image of orgs/movements pushing for AI x-safety"
X Link 2023-10-18T00:23Z [----] followers, [---] engagements
"Very happy to have supported this research at @MATSprogram. Applications for Summer [----] launching soon New Anthropic research: Auditing Language Models for Hidden Objectives. We deliberately trained a model with a hidden misaligned objective and put researchers to the test: Could they figure out the objective without being told https://t.co/fxmA9Os2C9 New Anthropic research: Auditing Language Models for Hidden Objectives. We deliberately trained a model with a hidden misaligned objective and put researchers to the test: Could they figure out the objective without being told"
X Link 2025-03-13T21:20Z [----] followers, [---] engagements
"New Anthropic research: Auditing Language Models for Hidden Objectives. We deliberately trained a model with a hidden misaligned objective and put researchers to the test: Could they figure out the objective without being told"
X Link 2025-03-13T16:07Z 837.8K followers, 354.9K engagements
"@AnthropicAI @GoogleDeepMind @OpenAI @AISecurityInst @RANDCorporation @redwood_ai @METR_Evals @apolloaievals @Atla_AI @TimaeusResearch @Leap_Labs @theoremlabs @WorkshopLabsPBC MATS connects researchers with world-class mentors including @sleepinyourhat Nicholas Carlini @NeelNanda5 @EthanJPerez @McaleerStephen @vkrakovna @yonashav @StephenLCasper @bshlgrs @MariusHobbhahn @RichardMCNgo and more @janleike @AlecRad etc. often collaborate as advisors"
X Link 2025-08-29T21:22Z [----] followers, [----] engagements
"@AnthropicAI @GoogleDeepMind @OpenAI @AISecurityInst @RANDCorporation @redwood_ai @METR_Evals @apolloaievals @Atla_AI @TimaeusResearch @Leap_Labs @theoremlabs @WorkshopLabsPBC @sleepinyourhat @NeelNanda5 @EthanJPerez @McaleerStephen @vkrakovna @yonashav @StephenLCasper @bshlgrs @MariusHobbhahn @RichardMCNgo @janleike @AlecRad Apply by Oct [--] midnight AoE Visit the @MATSProgram website for detailed information on the program and application process. http://matsprogram.org/apply http://matsprogram.org/apply"
X Link 2025-08-29T21:23Z [----] followers, [----] engagements
"Currently accepting AI safety research mentors for a Winter program; message me if you are interested Past mentors include: http://serimats.org/mentors http://serimats.org/mentors"
X Link 2023-09-05T23:05Z [----] followers, [----] engagements
"Also: - What moral personhood will digital minds have - How should nations share decision making power regarding world-transforming and Mercury-disassembling technology"
X Link 2024-07-07T00:29Z [----] followers, [----] engagements
"Big support to @austinc3301 and his new project AI safety student groups and entry-level internships are very important to the @MATSprogram pipeline (and all of AI safety). On more personal news I'm now the Co-Director of Kairos a new AI safety fieldbuilding org https://t.co/GAsLMmfSlY On more personal news I'm now the Co-Director of Kairos a new AI safety fieldbuilding org https://t.co/GAsLMmfSlY"
X Link 2024-10-25T23:13Z [----] followers, [----] engagements
"On more personal news I'm now the Co-Director of Kairos a new AI safety fieldbuilding org https://forum.effectivealtruism.org/posts/tTCyWthLyHk32WX9n/introducing-kairos-a-new-home-for-spar-and-fsp https://forum.effectivealtruism.org/posts/tTCyWthLyHk32WX9n/introducing-kairos-a-new-home-for-spar-and-fsp"
X Link 2024-10-25T22:10Z [----] followers, [----] engagements
"AI safety research that reduces the risk of non-catastrophic accidents or misuse (e.g. hate speech) makes commercial AI more viable driving AI hype and capabilities research. While important this research might fail to prevent genuinely catastrophic "black swan" risk"
X Link 2023-03-14T17:39Z [----] followers, [----] engagements
"Ever wanted to contribute to technical AI safety but haven't built a transformer Apply to The ML for AI safety bootcamp will run Sep 2-Oct [--] out of Applications close Jul [--]. http://SafeAI.org.uk http://ARENA.education http://SafeAI.org.uk http://ARENA.education"
X Link 2024-07-09T19:39Z [----] followers, [---] engagements
"@mealreplacer Good evening Robert"
X Link 2023-05-23T21:13Z [----] followers, [---] engagements
"MATS has accelerated 450+ researchers in the past [---] years. 80% of MATS alumni who graduated before [----] are working on AI safety/security including 200+ at @AnthropicAI @GoogleDeepMind @OpenAI @AISecurityInst @RANDCorporation @redwood_ai @METR_Evals @apolloaievals and more"
X Link 2025-08-29T21:18Z [----] followers, [----] engagements
"Our selection process is ongoing but it looks like around 75% of mentor applicants are rated above our minimum bar by our Mentor Selection Committee. If we accept [--] mentors that leaves [--] great mentors unsupported"
X Link 2025-06-02T23:05Z [----] followers, [----] engagements
"Precipissed: Feeling angry about civilizational inadequacy towards mitigating x-risk"
X Link 2022-12-15T20:14Z [----] followers, [---] engagements
"Excited to have supported this research @MATSprogram New paper on evaluating instrumental self-reasoning ability in frontier models 🤖🪞 We propose a suite of agentic tasks that are more diverse than prior work and give us a more representative picture of how good models are at eg. self-modification and embedded reasoning https://t.co/EM8X97MeBo New paper on evaluating instrumental self-reasoning ability in frontier models 🤖🪞 We propose a suite of agentic tasks that are more diverse than prior work and give us a more representative picture of how good models are at eg. self-modification and"
X Link 2024-12-06T19:19Z [----] followers, [---] engagements
"New paper on evaluating instrumental self-reasoning ability in frontier models 🤖🪞 We propose a suite of agentic tasks that are more diverse than prior work and give us a more representative picture of how good models are at eg. self-modification and embedded reasoning"
X Link 2024-12-06T16:37Z [----] followers, [----] engagements
"Some reasons you shouldn't assume civilization is adequate at solving AI alignment by default:"
X Link 2023-03-12T22:41Z [----] followers, [----] engagements
"Last chance to apply to work at MATS Still taking applications for Research Managers Community Managers and Operations Generalists. Apply by May [--] https://www.matsprogram.org/careers https://www.matsprogram.org/careers"
X Link 2025-04-30T23:30Z [----] followers, [----] engagements
"Last day to apply to @MATSprogram to help advance beneficial AI Last cohort scholars rated the program 9.2/10 on average (NPS: +74) and mentors advocated for scholars' research continuing at 8.1/10 on average (NPS: +25). Come see why https://airtable.com/appPxJ0QMqR7TElYU/pagRPwHQtcN8L0vIE/form https://airtable.com/appPxJ0QMqR7TElYU/pagRPwHQtcN8L0vIE/form"
X Link 2024-04-07T04:47Z [----] followers, [----] engagements
"@AnthropicAI @GoogleDeepMind @OpenAI @AISecurityInst @RANDCorporation @redwood_ai @METR_Evals @apolloaievals 10% of MATS alumni who graduated before [----] co-founded active AI safety/security start-ups including @apolloaievals @Atla_AI @TimaeusResearch @Leap_Labs @theoremlabs @WorkshopLabsPBC and more"
X Link 2025-08-29T21:19Z [----] followers, [----] engagements
"I just left a comment on @pibbssai's @manifund grant request (which I funded $25k) that AI safety people might find interesting. PIBBSS needs more funding https://manifund.org//projects/pibbss---affiliate-program-funding-6-months-6-affiliates-or-moretab=comments#7aa374d7-c42a-4519-9be2-08ccc03fed62 https://manifund.org//projects/pibbss---affiliate-program-funding-6-months-6-affiliates-or-moretab=comments#7aa374d7-c42a-4519-9be2-08ccc03fed62"
X Link 2024-09-15T19:22Z [----] followers, [---] engagements
"Good night Roberts"
X Link 2024-05-11T03:00Z [----] followers, [---] engagements
"@MATSprogram Alumni Impact Analysis published 78% of alumni are still working on AI alignment/control and 7% are working on AI capabilities. 68% have published alignment research https://www.lesswrong.com/posts/jeBkx6agMuBCQW94C/mats-alumni-impact-analysis https://www.lesswrong.com/posts/jeBkx6agMuBCQW94C/mats-alumni-impact-analysis"
X Link 2024-10-01T19:36Z [----] followers, [----] engagements
"AI alignment fieldbuilders often advocate a "hits-based" approach due to the "long tailed distribution of individual impact." But if IQ is normally distributed why is impact long-tailed My hypothesis "Luck": e.g. high-quality mentorship accessible problem framings financial freedom etc"
X Link 2023-07-04T19:49Z [----] followers, [----] engagements
"@MATSprogram is accepting applications for mentors in our Summer [----] Program. Please DM me if interested In addition to technical AI safety researchers we are interested in supporting AI gov infosec and natsec mentors"
X Link 2024-01-28T22:29Z [----] followers, [----] engagements
"I'm a @manifund Regrantor. I added some requests for funding proposals here: https://www.lesswrong.com/posts/tPjAgWpsQrveFECWP/ryan-kidd-s-shortformcommentId=uWwdHtsuLDDSJ9h9N https://www.lesswrong.com/posts/tPjAgWpsQrveFECWP/ryan-kidd-s-shortformcommentId=uWwdHtsuLDDSJ9h9N"
X Link 2024-05-25T21:51Z [----] followers, [----] engagements
"New MATS post on the current opportunities in technical AI safety as informed by [--] interviews with AI safety field leaders https://www.lesswrong.com/posts/QzQQvGJYDeaDE4Cfg/talent-needs-in-technical-ai-safety https://www.lesswrong.com/posts/QzQQvGJYDeaDE4Cfg/talent-needs-in-technical-ai-safety"
X Link 2024-05-25T02:40Z [----] followers, [---] engagements
"The majority of experts think AI catastrophic risk is worryingly high. Don't "Don't Look Up" btw i think it's totally possible that we're all just wrong about near-term x-risk. like through a combination of selection effects drinking the koolaid and mutually reinforcing each other's views we've worked ourselves into a panic over an implausible scenario (read to end) btw i think it's totally possible that we're all just wrong about near-term x-risk. like through a combination of selection effects drinking the koolaid and mutually reinforcing each other's views we've worked ourselves into a"
X Link 2025-02-01T00:00Z [----] followers, [---] engagements
"btw i think it's totally possible that we're all just wrong about near-term x-risk. like through a combination of selection effects drinking the koolaid and mutually reinforcing each other's views we've worked ourselves into a panic over an implausible scenario (read to end) like maybe there is a huge silent majority of AI experts who think our concerns are totally silly it's impossible to determine whether this is the case because people who are unworried about x-risk usually don't discuss it maybe alignment is trivially easy. maybe it turns out not to be a real problem. maybe we really"
X Link 2025-01-30T15:42Z [----] followers, 80.9K engagements
"Applications to Neel Nanda's Winter [----] @MATSprogram stream have launched My Winter MATS applications are open You'll work full-time writing a mech interp paper supervised by me. Due Aug [--] I've supervised 30+ papers by now (incl [--] top conference papers) but cohorts still get better each time. I'm hyped to see what this cohort achieves Highlights: https://t.co/pPoITAdl1A My Winter MATS applications are open You'll work full-time writing a mech interp paper supervised by me. Due Aug [--] I've supervised 30+ papers by now (incl [--] top conference papers) but cohorts still get better each time."
X Link 2025-07-30T00:49Z [----] followers, [----] engagements
"My Winter MATS applications are open You'll work full-time writing a mech interp paper supervised by me. Due Aug [--] I've supervised 30+ papers by now (incl [--] top conference papers) but cohorts still get better each time. I'm hyped to see what this cohort achieves Highlights:"
X Link 2025-07-29T23:54Z 36.4K followers, 116.3K engagements
"A snapshot of the AI safety community's research interests based on @MATSprogram applications. Note that MATS has historically had a technical AI safety focus and AI gov/policy + infosec interest might be underrepresented here"
X Link 2024-04-09T08:56Z [----] followers, [----] engagements
""In one hour the chatbots suggested four potential pandemic pathogens explained how they can be generated from synthetic DNA using reverse genetics supplied the names of DNA synthesis companies unlikely to screen orders." https://arxiv.org/abs/2306.03809 https://arxiv.org/abs/2306.03809"
X Link 2023-07-19T02:01Z [----] followers, [----] engagements
"@AnthropicAI @GoogleDeepMind @OpenAI @AISecurityInst @RANDCorporation @redwood_ai @METR_Evals @apolloaievals @Atla_AI @TimaeusResearch @Leap_Labs @theoremlabs @WorkshopLabsPBC @sleepinyourhat @NeelNanda5 @EthanJPerez @McaleerStephen @vkrakovna @yonashav @StephenLCasper @bshlgrs @MariusHobbhahn @RichardMCNgo @janleike @AlecRad Participants rated our last program 9.4/10 on average with a median of 10/10 75/98 researchers are continuing in our 6-month extension program. All nationalities are eligible to participate in MATS; 50% of our scholars are international"
X Link 2025-08-29T21:22Z [----] followers, [----] engagements
"Last week to apply to @NeelNanda5's mechanistic interpretability MATS stream Applications close Nov [--] 11:59 pm PT. http://matsprogram.org/interpretability http://matsprogram.org/interpretability"
X Link 2023-11-04T23:16Z [----] followers, [----] engagements
"Another excellent (and disturbing) paper from Owain in collaboration with @MATSprogram Surprising new results: We finetuned GPT4o on a narrow task of writing insecure code without warning the user. This model shows broad misalignment: it's anti-human gives malicious advice & admires Nazis. This is emergent misalignment & we cannot fully explain it 🧵 https://t.co/kAgKNtRTOn Surprising new results: We finetuned GPT4o on a narrow task of writing insecure code without warning the user. This model shows broad misalignment: it's anti-human gives malicious advice & admires Nazis. This is"
X Link 2025-02-25T19:16Z [----] followers, [---] engagements
"Surprising new results: We finetuned GPT4o on a narrow task of writing insecure code without warning the user. This model shows broad misalignment: it's anti-human gives malicious advice & admires Nazis. This is emergent misalignment & we cannot fully explain it 🧵"
X Link 2025-02-25T17:17Z 17.8K followers, 1.9M engagements
"I propose a new name for an important metaethical distinction: bosonic vs. fermionic moral theories. Bosons are particles that can degenerately occupy the same state while fermions can only occupy individual states"
X Link 2025-06-29T18:28Z [----] followers, [----] engagements
"@MATSprogram @NeelNanda5 @EthanJPerez @OwainEvans_UK @EvanHub @bshlgrs @dawnsongtweets @DavidSKrueger @RichardMCNgo Other @MATSprogram mentors include Nicholas Carlini @McaleerStephen @achan96 @ben_s_bucknall @MichaelD1729 @FlorianTramer @SamuelAlbanie @davlindner @Turn_Trout @emmons_scott @MrinankSharma and many more https://matsprogram.org/mentors https://matsprogram.org/mentors"
X Link 2025-03-20T20:02Z [----] followers, [---] engagements
"I did a podcast Thanks again for having me on @soroushjp; it was a lot of fun https://x.com/soroushjp/status/1722336164793962603s=20 📣 EP10 AGI Show w/ @ryan_kidd44 out We talk ML Alignment & Theory Scholars (MATS) program that accelerates people into AI safety research roles via mentorship seminars & connections. If you're interested in technical AI research for catastrophic/x-risk this ep is for you https://t.co/qPTscUNdId https://x.com/soroushjp/status/1722336164793962603s=20 📣 EP10 AGI Show w/ @ryan_kidd44 out We talk ML Alignment & Theory Scholars (MATS) program that accelerates people"
X Link 2023-11-08T23:21Z [----] followers, [----] engagements
"📣 EP10 AGI Show w/ @ryan_kidd44 out We talk ML Alignment & Theory Scholars (MATS) program that accelerates people into AI safety research roles via mentorship seminars & connections. If you're interested in technical AI research for catastrophic/x-risk this ep is for you"
X Link 2023-11-08T19:31Z [----] followers, [----] engagements
"@AnthropicAI @GoogleDeepMind @OpenAI @AISecurityInst @RANDCorporation @redwood_ai @METR_Evals @apolloaievals @Atla_AI @TimaeusResearch @Leap_Labs @theoremlabs @WorkshopLabsPBC In [---] years MATS researchers have coauthored 115+ arXiv papers with 5100+ citations and an org h-index of [--]. We are experts at accelerating awesome researchers with mentorship compute support and community"
X Link 2025-08-29T21:19Z [----] followers, [----] engagements
"Happy to have supported this research @MATSprogram AI Governance should work with markets not against them Excited to finally share a preprint that @FranklinMatija @rupal15081 & I have been working on. https://t.co/5XDxVCQvaN AI Governance should work with markets not against them Excited to finally share a preprint that @FranklinMatija @rupal15081 & I have been working on. https://t.co/5XDxVCQvaN"
X Link 2025-02-05T18:32Z [----] followers, [---] engagements
"AI Governance should work with markets not against them Excited to finally share a preprint that @FranklinMatija @rupal15081 & I have been working on"
X Link 2025-02-05T16:01Z [---] followers, 18.8K engagements
"If the current race towards AGI worries you come work on AI safety The field is highly impactful talent constrained and filled with low-hanging fruit. https://80000hours.org/problem-profiles/artificial-intelligence/#what-can-you-do-concretely-to-help https://80000hours.org/problem-profiles/artificial-intelligence/#what-can-you-do-concretely-to-help"
X Link 2024-12-21T22:59Z [----] followers, [----] engagements
"So excited to see @apolloaisafety launch https://x.com/apolloaisafety/status/1663582940658270210 Hi we are Apollo Research-a new AI evals research organization. Our research agenda is focused on interpretability and behavioral model evaluations. We intend to apply our findings and cooperate with AI labs to prevent the deployment of deceptive AIs https://t.co/lcvyGNJg3w https://x.com/apolloaisafety/status/1663582940658270210 Hi we are Apollo Research-a new AI evals research organization. Our research agenda is focused on interpretability and behavioral model evaluations. We intend to apply our"
X Link 2023-05-30T17:18Z [----] followers, [---] engagements
"Hi we are Apollo Research-a new AI evals research organization. Our research agenda is focused on interpretability and behavioral model evaluations. We intend to apply our findings and cooperate with AI labs to prevent the deployment of deceptive AIs https://www.apolloresearch.ai/blog/announcement https://www.apolloresearch.ai/blog/announcement"
X Link 2023-05-30T16:27Z [----] followers, 29.6K engagements
"The MATS Winter 2023-24 Retrospective is published https://www.lesswrong.com/posts/Z87fSrxQb4yLXKcTk/mats-winter-2023-24-retrospective https://www.lesswrong.com/posts/Z87fSrxQb4yLXKcTk/mats-winter-2023-24-retrospective"
X Link 2024-05-11T21:05Z [----] followers, [----] engagements
"Excited to have helped support this documentary via the @manifund regranting program 'Regulation shouldn't be written in blood.' My documentary on California's most controversial AI bill SB-1047 is finally out on Youtube. Go watch it https://t.co/06iUovQNw4 'Regulation shouldn't be written in blood.' My documentary on California's most controversial AI bill SB-1047 is finally out on Youtube. Go watch it https://t.co/06iUovQNw4"
X Link 2025-05-02T23:21Z [----] followers, [---] engagements
"'Regulation shouldn't be written in blood.' My documentary on California's most controversial AI bill SB-1047 is finally out on Youtube. Go watch it"
X Link 2025-05-02T21:16Z 19.4K followers, 24.9K engagements
"If income is lognormally distributed and happiness is logarithmic in wealth then happiness is normally distributed in the US"
X Link 2024-06-15T22:56Z [----] followers, [----] engagements
"@MATSprogram is now hiring for a Research Manager role based in London Come help us grow the AI safety research field :) https://www.matsprogram.org/careers https://www.matsprogram.org/careers"
X Link 2024-07-09T19:23Z [----] followers, [----] engagements
"@thlarsen What do you think the holes are"
X Link 2025-06-29T01:56Z [----] followers, [---] engagements
"@BogdanIonutCir2 Not even that Our funders are incredibly supportive. MATS is constrained on organization capacity and experience not scholars mentors or funding. We have recently hired [--] (soon 17) new staff effectively doubling in size. Hopefully growing the Program team soon too"
X Link 2025-06-03T17:53Z [----] followers, [---] engagements
"@Pandora_Delaney @So8res @Aella_Girl @JamieWahls @asteriskmgzn Unsong"
X Link 2025-03-22T17:07Z [----] followers, [---] engagements
"And that's just for our excellence bar; 97% of mentor applicants were above our selection committee's indifference point"
X Link 2025-06-03T18:26Z [----] followers, [---] engagements
"@GaryMarcus Please do it. LessWrong has its flaws but it's still the best forum for AI futurism and you bring an important perspective"
X Link 2025-06-29T02:09Z [----] followers, [---] engagements
"I recently gave a talk to the AI Alignment Network in Japan about our work at @MATSprogram. Recording of ALIGN Webinar #4 with Dr. Ryan Kidd is now available Here @ryan_kidd44 provided a very accessible explanation of what AGI risks are what countermeasures are needed for different scenarios how MATS is addressing its talent needs. https://t.co/QyTsEJF8gE Recording of ALIGN Webinar #4 with Dr. Ryan Kidd is now available Here @ryan_kidd44 provided a very accessible explanation of what AGI risks are what countermeasures are needed for different scenarios how MATS is addressing its talent needs."
X Link 2024-06-23T19:25Z [----] followers, [----] engagements
"Recording of ALIGN Webinar #4 with Dr. Ryan Kidd is now available Here @ryan_kidd44 provided a very accessible explanation of what AGI risks are what countermeasures are needed for different scenarios how MATS is addressing its talent needs. https://www.youtube.com/watchv=-NiUg8QSBcM https://www.youtube.com/watchv=-NiUg8QSBcM"
X Link 2024-06-23T10:17Z [---] followers, [----] engagements
"I'd love to support this research with funding http://Manifund.org http://Manifund.org"
X Link 2024-07-07T00:33Z [----] followers, [---] engagements
"Fermionic moral theories value new moral patients only insofar as they have different experiences. "Moral degeneracy pressure" would disfavor the creation of identical copies as they would be treated like "pointers" to the original rather than independent moral patients. Under these theories inequality and maybe even some suffering entities are permissible if higher value states are already occupied by other entities. A "fermionic moral utopia" could look like the universe filled with minds experiencing infinitesimally varying distinct positive experiences"
X Link 2025-06-29T18:29Z [----] followers, [---] engagements
"@sama "Leading the pack from the front" likely requires selling shares to Moloch. Exercising caution might require buying them back (hard) or dropping out. If AGI is imminent this might not matter but I'm not sure it is"
X Link 2023-11-19T03:36Z [----] followers, [----] engagements
"MATS is hiring Research Managers Community Managers and Operations Generalists Rolling applications close May [--]. Come align and secure AI with us https://www.matsprogram.org/careers https://www.matsprogram.org/careers"
X Link 2025-04-05T17:13Z [----] followers, [---] engagements
"Some takeaways from a recent conference that discussed AI safety:"
X Link 2023-03-10T20:24Z [----] followers, [----] engagements
"Existential hope Incoming Commerce Secretary Lutnick on AI export controls at confirmation hearing: "AI chip smuggling has got to end" We need to "stop giving them our tools so they can compete with us" "I'm thrilled to empower BIS" We're so incredibly back https://t.co/WVJYHx8IIt Incoming Commerce Secretary Lutnick on AI export controls at confirmation hearing: "AI chip smuggling has got to end" We need to "stop giving them our tools so they can compete with us" "I'm thrilled to empower BIS" We're so incredibly back https://t.co/WVJYHx8IIt"
X Link 2025-01-30T19:41Z [----] followers, [---] engagements
"Incoming Commerce Secretary Lutnick on AI export controls at confirmation hearing: "AI chip smuggling has got to end" We need to "stop giving them our tools so they can compete with us" "I'm thrilled to empower BIS" We're so incredibly back"
X Link 2025-01-29T18:09Z [----] followers, 32.6K engagements
"LISA (@LondonSafeAI) is hiring a CEO The LISA office is home to @apolloaisafety @BlueDotImpact @MATSprogram extension and other top-tier AI safety projects Apps due Feb [--]. https://london-safe-ai.notion.site/chiefexecutiveofficer https://london-safe-ai.notion.site/chiefexecutiveofficer"
X Link 2025-02-09T19:31Z [----] followers, [----] engagements
"MATS mentors for Winter 2024-25 include @bshlgrs @EthanJPerez @NeelNanda5 @OwainEvans_UK @eli_lifland @DKokotajlo67142 @EvanHub @StephenLCasper @FabienDRoger @seb_far @Turn_Trout @davlindner @fiiiiiist @MrinankSharma @DavidSKrueger @leedsharkey @SamuelAlbanie and more"
X Link 2024-09-12T21:58Z [----] followers, [---] engagements
Limited data mode. Full metrics available with subscription: lunarcrush.com/pricing