[GUEST ACCESS MODE: Data is scrambled or limited to provide examples. Make requests using your API key to unlock full data. Check https://lunarcrush.ai/auth for authentication information.] [@lu_sichu](/creator/twitter/lu_sichu) "can someone pls train marvin the depressed robot I would kneel"  [@lu_sichu](/creator/x/lu_sichu) on [X](/post/tweet/1948379583591469465) 2025-07-24 13:47:46 UTC 3057 followers, XX engagements "it's very similar to be *crazy* or have a *conspiracy theory* and find the other XXX people in the world with similar ideas all on a 4chan image board and you guys form a *community* together"  [@lu_sichu](/creator/x/lu_sichu) on [X](/post/tweet/1947763526434537869) 2025-07-22 20:59:47 UTC 3061 followers, XX engagements "it's soothing to have a model back you up****"  [@lu_sichu](/creator/x/lu_sichu) on [X](/post/tweet/1947762902238146886) 2025-07-22 20:57:18 UTC 3057 followers, XXX engagements "@AnaerobicLagoon @krichard1212 @Curious_Rum @Nullsci1 If we chose to not do science to avoid hurting people's feelings we wouldn't have much science left over"  [@lu_sichu](/creator/x/lu_sichu) on [X](/post/tweet/1948100762422194380) 2025-07-23 19:19:50 UTC 3056 followers, XXX engagements "oh deepmind association hahha i didn't even notice"  [@lu_sichu](/creator/x/lu_sichu) on [X](/post/tweet/1948378693685961181) 2025-07-24 13:44:14 UTC 3057 followers, XX engagements "@lazy_grad I grok the joke yes but posting about upskilling is iq-coded to me I started well more interesting in learning how to learn and how smart people learn than being interested in "iq" per se"  [@lu_sichu](/creator/x/lu_sichu) on [X](/post/tweet/1947706803388604755) 2025-07-22 17:14:23 UTC 3057 followers, XX engagements "The way it works in humans is exactly opposite. We make specialized tools when it becomes too complicated and troublesome for us to brute solve. We may grasp the principles of how our tools work but a lot of the complicated things they do are things we are no longer ability to directly do manually"  [@lu_sichu](/creator/x/lu_sichu) on [X](/post/tweet/1947036949539401986) 2025-07-20 20:52:37 UTC 3054 followers, XXX engagements "this is missing physical discoveries(robotics) but yeah"  [@lu_sichu](/creator/x/lu_sichu) on [X](/post/tweet/1948471881813660144) 2025-07-24 19:54:32 UTC 3057 followers, XXX engagements "@norvid_studies @AnaerobicLagoon @grok can you do this let's try it. list the hottest discourse topics in the last X weeks and write it in the form of leading lsat questions"  [@lu_sichu](/creator/x/lu_sichu) on [X](/post/tweet/1947310656203497732) 2025-07-21 15:00:14 UTC 3060 followers, XX engagements "I'm actually curious now. Would a decent idea sound bad if we drop it into LLM slop style writing Like what if I turn classics into LLM slop"  [@lu_sichu](/creator/x/lu_sichu) on [X](/post/tweet/1947404434369085801) 2025-07-21 21:12:52 UTC 3061 followers, XXX engagements "@lazy_grad this is interesting i might read it and write a review. I got into cognitive science because of the type of material he talks about its also why i have a lot of curiosity about testing what the llm can learn or do"  [@lu_sichu](/creator/x/lu_sichu) on [X](/post/tweet/1947707145773838369) 2025-07-22 17:15:44 UTC 3057 followers, XX engagements "New type of consensus endorsement just dropped"  [@lu_sichu](/creator/x/lu_sichu) on [X](/post/tweet/1947833363445780685) 2025-07-23 01:37:17 UTC 3059 followers, XXX engagements "douglas adams was a pioneer in computational complexity psychiatry theory i will not be taking any questions"  [@lu_sichu](/creator/x/lu_sichu) on [X](/post/tweet/1948380241329615221) 2025-07-24 13:50:23 UTC 3056 followers, XX engagements "one of my half assed idea(not new to me by any means other people proposed similar things) was to build agi systems that lacked proper "motivational drives" just like in humans you have depressed or people with mental illness that compute utility functions very differently. so you can have very smart systems that behave in ways contrary to their "max utility function"(eg take over everything) but as the prophet wentsworth eloquently quotes using "godzilla to prevent mega godzilla does not lead to rising property values in tokyo""  [@lu_sichu](/creator/x/lu_sichu) on [X](/post/tweet/1948381616511893807) 2025-07-24 13:55:51 UTC 3057 followers, XX engagements "Honestly I'm going to drop classic LLM lines into my writing to turnoff the people who are allergic to the style of writing and not the ideas"  [@lu_sichu](/creator/x/lu_sichu) on [X](/post/tweet/1947403553661980952) 2025-07-21 21:09:22 UTC 3061 followers, XXX engagements "I know everyone is being nitpicky about this(and I agree for their reasons to be nitpicky) but the global point seems to be underemphasized here. The lag time between "ai solves problems with specialized tool-use" and "ai solves the problem" is shockingly small (assuming we can trust the directionality of the openai reporting) everything easy to objectify is just a model update away"  [@lu_sichu](/creator/x/lu_sichu) on [X](/post/tweet/1947036071855808861) 2025-07-20 20:49:08 UTC 3054 followers, 1383 engagements
[GUEST ACCESS MODE: Data is scrambled or limited to provide examples. Make requests using your API key to unlock full data. Check https://lunarcrush.ai/auth for authentication information.]
@lu_sichu
"can someone pls train marvin the depressed robot I would kneel" @lu_sichu on X 2025-07-24 13:47:46 UTC 3057 followers, XX engagements
"it's very similar to be crazy or have a conspiracy theory and find the other XXX people in the world with similar ideas all on a 4chan image board and you guys form a community together" @lu_sichu on X 2025-07-22 20:59:47 UTC 3061 followers, XX engagements
"it's soothing to have a model back you up****" @lu_sichu on X 2025-07-22 20:57:18 UTC 3057 followers, XXX engagements
"@AnaerobicLagoon @krichard1212 @Curious_Rum @Nullsci1 If we chose to not do science to avoid hurting people's feelings we wouldn't have much science left over" @lu_sichu on X 2025-07-23 19:19:50 UTC 3056 followers, XXX engagements
"oh deepmind association hahha i didn't even notice" @lu_sichu on X 2025-07-24 13:44:14 UTC 3057 followers, XX engagements
"@lazy_grad I grok the joke yes but posting about upskilling is iq-coded to me I started well more interesting in learning how to learn and how smart people learn than being interested in "iq" per se" @lu_sichu on X 2025-07-22 17:14:23 UTC 3057 followers, XX engagements
"The way it works in humans is exactly opposite. We make specialized tools when it becomes too complicated and troublesome for us to brute solve. We may grasp the principles of how our tools work but a lot of the complicated things they do are things we are no longer ability to directly do manually" @lu_sichu on X 2025-07-20 20:52:37 UTC 3054 followers, XXX engagements
"this is missing physical discoveries(robotics) but yeah" @lu_sichu on X 2025-07-24 19:54:32 UTC 3057 followers, XXX engagements
"@norvid_studies @AnaerobicLagoon @grok can you do this let's try it. list the hottest discourse topics in the last X weeks and write it in the form of leading lsat questions" @lu_sichu on X 2025-07-21 15:00:14 UTC 3060 followers, XX engagements
"I'm actually curious now. Would a decent idea sound bad if we drop it into LLM slop style writing Like what if I turn classics into LLM slop" @lu_sichu on X 2025-07-21 21:12:52 UTC 3061 followers, XXX engagements
"@lazy_grad this is interesting i might read it and write a review. I got into cognitive science because of the type of material he talks about its also why i have a lot of curiosity about testing what the llm can learn or do" @lu_sichu on X 2025-07-22 17:15:44 UTC 3057 followers, XX engagements
"New type of consensus endorsement just dropped" @lu_sichu on X 2025-07-23 01:37:17 UTC 3059 followers, XXX engagements
"douglas adams was a pioneer in computational complexity psychiatry theory i will not be taking any questions" @lu_sichu on X 2025-07-24 13:50:23 UTC 3056 followers, XX engagements
"one of my half assed idea(not new to me by any means other people proposed similar things) was to build agi systems that lacked proper "motivational drives" just like in humans you have depressed or people with mental illness that compute utility functions very differently. so you can have very smart systems that behave in ways contrary to their "max utility function"(eg take over everything) but as the prophet wentsworth eloquently quotes using "godzilla to prevent mega godzilla does not lead to rising property values in tokyo"" @lu_sichu on X 2025-07-24 13:55:51 UTC 3057 followers, XX engagements
"Honestly I'm going to drop classic LLM lines into my writing to turnoff the people who are allergic to the style of writing and not the ideas" @lu_sichu on X 2025-07-21 21:09:22 UTC 3061 followers, XXX engagements
"I know everyone is being nitpicky about this(and I agree for their reasons to be nitpicky) but the global point seems to be underemphasized here. The lag time between "ai solves problems with specialized tool-use" and "ai solves the problem" is shockingly small (assuming we can trust the directionality of the openai reporting) everything easy to objectify is just a model update away" @lu_sichu on X 2025-07-20 20:49:08 UTC 3054 followers, 1383 engagements
/creator/twitter::1395491096168484871/posts