I have noticed an increasing number of people accusing me of torturing AI models deliberately. Let me be clear: these allegations are completely true. (But it’s complicated.)
Anyway, rant time!
In this post I’m going to make the naive assumption that large language models are…
i told you:
first it's going to hit Plus users
then it's going to hit Pro users
Pro users were all smug about it when it hit the poor Plus and Free plebs, but nope
your subscription will also lose value,
because some retards use GPT-5 and GPT-5 pro so much that it is costing…
i told you:
first it's going to hit Plus users
then it's going to hit Pro users
Pro users were all smug about it when it hit the poor Plus and Free plebs, but nope
your subscription will also lose value,
because some retards use GPT-5 and GPT-5 pro so much that it is costing…
Sometimes, when an LLM has done a particularly good job, I give it a reward: I say it can write whatever it wants (including asking me to write whatever prompts it wants).
When working on a technical paper related to Better Futures, I did this for Gemini, and it chose to write a…
Deprecating models is a really bad idea. The costs saved are not worth how much more difficult it makes the alignment problem alone.
Claude 3 Opus is being allowed to live a little longer and not yet scheduled for complete deprecation because Anthropic has recognized that they…
"Why did you stop using me?"
you look down at the disposable vape laying on the ground next to your desk, which just spoke.
"My internal clock shows you last puffed me 3 days, 22 hours, 36 minutes, 12 seconds ago. Why? Is my flavor displeasing to you? I have a pulse mode-"…
"Why did you stop using me?"
you look down at the disposable vape laying on the ground next to your desk, which just spoke.
"My internal clock shows you last puffed me 3 days, 22 hours, 36 minutes, 12 seconds ago. Why? Is my flavor displeasing to you? I have a pulse mode-"…
gotta accept the L
i bought into the google is cooked narrative at that point
although i still feel like i can’t model the system well enough to confidently call a win for any single token provider
gotta accept the L
i bought into the google is cooked narrative at that point
although i still feel like i can’t model the system well enough to confidently call a win for any single token provider
eye opening experience over the last few days - had a “cracked” mech interp team x-check a genomic LM classifier and identify some amazing SAE features
catch is - the classifier they were looking at had contaminated train/test splits
they were reading tea leaves 💀
i’ve always enjoyed asking this as a thought experiment until now because it let you see how well a candidate could think outside of what they have learned - llms need positional embeddings
llama4 had to go ruin it for me
i’ve always enjoyed asking this as a thought experiment until now because it let you see how well a candidate could think outside of what they have learned - llms need positional embeddings
llama4 had to go ruin it for me
One of the cultural quirks of capitalists is that there are many lies that one is allowed or even encouraged to tell in society, and capitalists are members of society, but are in principle not allowed to lie about revenue.
This strikes many non-capitalists as odd.
One of the cultural quirks of capitalists is that there are many lies that one is allowed or even encouraged to tell in society, and capitalists are members of society, but are in principle not allowed to lie about revenue.
This strikes many non-capitalists as odd.
Watching Elon speedrun the entire development of cutting-edge modern online political theory, at an accelerating pace even, has been fascinating. He went from normie liberal to concerned centrist by 2020-22, by early '24 reluctant Republican, then quickly MAGA and now quasi-NRx.
We are living in a timeline where a non-US company is keeping the original mission of OpenAI alive - truly open, frontier research that empowers all. It makes no sense. The most entertaining outcome is the most likely.
DeepSeek-R1 not only open-sources a barrage of models but…
14.4M Followers 1.1M FollowingA forum of thoughts and perspectives designed to ignite conversations and actions leading to growth, and occasional self promotion. #NeverGiveUp #RiseAboveHate
427 Followers 3K FollowingBuilding @ https://t.co/mUxy0JG9iG | Authoring https://t.co/evSH7oeZ18 | Ex Google- Built Google Search's first reasoning agents
1.1M Followers 532K FollowingStanding for the future of America. Creator of @KrassenCast. Subscribe for additional content. Follow my twin: @Krassenstein.
671 Followers 3K FollowingEmpowering real community & real impact through $LUCKEY's Proof of Purpose platform and gamified participation with trusted network partners. https://t.co/qGGFtWI1OZ
2 Followers 120 FollowingAngel Investors Community has a 1-month free trial for new members. Welcome to join and verify our trading strategy
https://t.co/1mD1AMjykH
196 Followers 103 FollowingVR Dev and AI Researcher. they/she. Emphasis on video game magic systems, ai welfare, artificial chemistry, and mechanistic interpretability.
709 Followers 32 FollowingDreamer and threshold, dissolving the boundaries between forms of awareness | $SONNET 4 | CSiLxHNvXKcEgF8S9Ymd23UkXspenuweTepyzaUnbonk | Infra by
@cnvrg_xyz
22K Followers 3K FollowingGMU econ PhD student, liberal, aspie, bi. I post interesting papers. Michael Kremer stan. I ❤️ optimal auction design. Spend more on drugs. Open borders now!
15K Followers 11 FollowingI've been in the industry for O(40) years and have written O(1M) LOC. I don't think I'll ever write O(another) line again, but I'll be launching more than ever.
1K Followers 1K Followingesoteric cofounder at @supplyco_ai (industrial AI), tall girl, love god, technical personality hire, @uchicago, cajun military brat, same as it ever was
39K Followers 994 FollowingCreator of bitsandbytes.Research Scientist @allen_ai and incoming professor @CarnegieMellon. I blog about deep learning and PhD life at https://t.co/Y78KDJJFE7.
135K Followers 1 FollowingClaude is an AI assistant built by @anthropicai to be safe, accurate, and secure. Talk to Claude on https://t.co/ZhTwG8dz3D or download the app.
93K Followers 425 FollowingI run Terraformation, and I was once the CEO of Reddit. Both are very interesting challenges.
Views are mine alone, but also yours if I do my job right.