Sam Ringer @sam_ringer
AI could be really, really bad. I'm trying to make it less bad at @AnthropicAI. Joined February 2012-
Tweets18
-
Followers156
-
Following147
-
Likes52
😲
This is one of the best posts I have read in a while. Multiple moments where the muddy waters of alignment became slightly clearer: alignmentforum.org/posts/vJFdjigz…
On watching the Robert Gadling sequence in The Sandman, I was struck by how little things changed between 1389 and 1889 (500 years!). Jump-cut to 1989 and BAM everything is different: phones, cars, speedboats... What the hell is 2089 going to look like when they meet next?!
You know you're in too deep with EA when your friends use "moral realist" as an insult
Wow, 2022 has been a pretty wild year in AI developments. Time to reset the clock.
Wow, 2022 has been a pretty wild year in AI developments. Time to reset the clock. https://t.co/fpEqGzS0D1
Our gift to the incoming @Speechmatics ML Engineers has arrived: @TomChivers's fantastically accessible introduction to AI Safety and rationality. (Managed to sneak my own copy in as well....)
It's also interesting how strong the evidence is on spaced repetition vs how it's a) almost totally UNdiscussed in education policy world & b) almost no teachers know about it & few schools teach it. My memory is shockingly bad, Anki is like magic
It's also interesting how strong the evidence is on spaced repetition vs how it's a) almost totally UNdiscussed in education policy world & b) almost no teachers know about it & few schools teach it. My memory is shockingly bad, Anki is like magic
I recently wrote a summary of @AnthropicAI's first paper: lesswrong.com/posts/oBpebs5j… It's really nice to see a lab working on Safety and not open-sourcing capability.
Markov blankets becoming more relevant than ever!
VQ-VAE can be interpreted as Bayesian Networks with Mixture-Gaussian Prior, if we treat l2 distance between z_q and z_e as logit. Interesting. arxiv.org/abs/2002.08111 This also reminds me of @kchonyc's recent post about Soft K-NN.
Aaditya ; @Aaditya26082004
524 Followers 7K Following CS'26 • Machine Learning • Open-Source • Web Dev. • Algorithms • Jai Shree Krishna 🦚🪈Abby Novick Hoskin @CorpusCalosseum
661 Followers 825 Following PhD in Psych/Neuro from Princeton. 80,000 Hours Advisor. Mother of two multimodal multitasking neural networks. Views are my own.Machel Reid @machelreid
2K Followers 1K Following Research Scientist @GoogleDeepMind Working on LLMs on the Gemini Team; did gemini 1.5 proNeil Houlsby @neilhoulsby
4K Followers 318 Following Professional AI researcher; amateur athlete. Senior Staff RS in the Google Deepmind, Zürich. Attempts triathlons.Pearl_US_ @PearlUS185773
21 Followers 3K Following I wish for what I want, I wish for what I want, and what I see when I look up is tenderness.Brandon McKinzie @mckbrando
2K Followers 2K Following Multimodal LLMs @Apple. Prev: Physics/CS @UCBerkeley.Afroz Mohiuddin @afrozenator
1K Followers 5K Following Research Engineer at Google Brain. Interested in Science, Psychology, Investing, Design and generally almost everything. Good Thoughts, Good Words, Good Deeds.Garrett -DeepWriterAI @DeepAIWriter
12K Followers 6K Following Over-engineering Agentic Systems for long-form writing. Generating scripts, fiction or non, breakthrough ideas, whole universes, etc. The Deep Writer. DM4demo.Ted Moskovitz @ted_moskovitz
740 Followers 192 Following PhD student at @GatsbyUCL. Formerly: intern at @DeepMind, @UberAILabs, student at @ColumbiaCompSci, @PrincetonNeuro.Yury Sulsky @ysulsky
96 Followers 114 FollowingAbhimanyu Sangitrao @abhi_manyu047
0 Followers 4 FollowingCristiano Giardina @CrisGiardina
1K Followers 3K Following Writing mostly about AI · "That most limited of all specialists, the well-rounded man."Sci.0 @ScientistZer0
3 Followers 128 Following Paper: Fundamental Logical Errors in Science and New Discoveries Changing Philosophy, Physics, Chemistry, Biology With GPT4s opinion on the paper (AI's review)123 @tttzzz000
114 Followers 290 FollowingImminent Eschatology @wmgcbr
40 Followers 621 Followingnick nassuphis @NNassuphis
119 Followers 5K FollowingBurny — Effective O.. @burny_tech
14K Followers 6K Following Transhuman engineer in singularity! Lover of AI & omnidisciplionary metamathemagics! Hypercuriousia! Omniperspectivity! Shapeshifting metafluid! Freedom 4 all!Langchain Philosopher @pranavmarla
358 Followers 2K Following Building truth seeking reasoning frameworks using LLMs // @Northeastern Created DebateTree and Recursive Reasoning. Looking for full-time work/internshipsSimon Geisler @geisler_si
582 Followers 215 Following Reliable and trustworthy machine learning on graphs. Computer Science PhD Student at @TU_MuenchenJack White @vzjgsbnrzs
29 Followers 229 FollowingNikhila Ravi @nikhilaravi
5K Followers 2K Following Research Engineer @AIatMeta (FAIR), @Cambridge_Uni, @kennedyscholars @harvard, @MCCOfficial cricketer 🇮🇳 🇬🇧 🇺🇸Josh Walker @Jwalks89
93 Followers 243 FollowingMichael Dempsey @mhdempsey
25K Followers 4K Following Dreaming about the future, romanticizing the past | MP @compoundvc investing in former science projects & crypto with a thesis-driven approach at seedOliver Parish @OllieParish
0 Followers 3 FollowingLiangyu Chen @cliangyu_
524 Followers 1K Followingtime traveling sikkun.. @254c_remove
641 Followers 5K Following TIMEWAR GOT YOU FEELING WEAK? GET OUR FREE 30-DAY BEACH BODY BOOTCAMP NOW! MAKE THE DARK FOREST YOUR STAGE WITH A NEW TRANSFORMATION!Jeremy Peterson @BoomBrusher
1K Followers 5K Following Automate the Neighborhood! Abundance on my block. How can we leverage AI in our neighborhoods to improve quality of life right now?Yao Fu @Francis_YAO_
14K Followers 2K Following PhD @EdinburghNLP on LLMs and Machine Reasoning. Ex. @Columbia @PKU1898 @MITIBMLab @allen_ai AGI has yet to come, so keep runningMd Anzar Ahmad @saad_anzar
54 Followers 639 Following Founder @__paperbrain. erevald at @_buildspace. I believe. In all of it. And more. 🕊️ilya vinnikov ❤️�.. @iliushavinnikov
224 Followers 449 Following Prof. @vinnikovLAB 🧬🐁🧪 | Born in Ukraine | Russian by birth | European by choice | Husband of 🎹https://t.co/JTAcD8e9a2 | Dad of 🎨@mischavinnikov | 🎼⚽️🏐 | AlignAGI❗Jonathan Mannhart is .. @JMannhart
2K Followers 1K Following Interested in: cognitive science, Bayes, (ir)rationality, (effective) altruism, happiness, AI Alignment, reward hacking, running, reading booksEJT @ejjiott
481 Followers 3K FollowingRobert Scoble @Scobleizer
504K Followers 68K Following Follow me on my new podcast with AI startups, Unaligned. Tech industry color commentator since 1993. Author/Blogger. Former strategist @Microsoft.Luke Harries @LukeHarries_
3K Followers 4K Following Head of Growth at @ElevenLabsio. Previously: Interim Head of Product @PostHog, Co-Founder @FellaHealth (backed by YC), ML Engineer @MicrosoftKelly Marchisio (St. .. @cheeesio
1K Followers 558 Following Member of Technical Staff @cohere. Formerly: PhD @jhuclsp Alexa Fellow @amazon dev @Google MPhil @cambridgenlp EdM @hgse 🔑🔑¬🧀 (@kelvenmar20)Trenton Bricken @TrentonBricken
6K Followers 2K Following Trying to figure out what makes minds and machines go "Beep Bop!" @AnthropicAIRichie @choose_richie
0 Followers 1K FollowingFS @Franksntra3
612 Followers 5K Following Finance, History , Religion , Geopolitics, Bio Gene , Anthro/ Paleontology and nothing more nothing lessYang Fan 范阳 @Yang_Supertramp
881 Followers 5K Following Will visit the US. in 2024. Figuring out AI x Bio intelligence, building sth new. ❤️ 🤖️🧬 🧠 🧘 🌍 Member of The Explorers Club.Sander Dieleman @sedielem
50K Followers 2K Following Research Scientist at Google DeepMind. I tweet about deep learning (research + software), music, generative models (personal account).orthonormal @renormalized
464 Followers 102 Following He/him. Don't follow me hoping for profundity. I have enough outlets for that in my life.Katja Grace 🔍 @KatjaGrace
8K Followers 798 Following Thinking about whether AI will destroy the world at https://t.co/pMilDvd4ya. DM or email for media requests. Feedback: https://t.co/zGAm1i7SKHBen Kuhn @benskuhn
7K Followers 289 Following Care a lot and try hard • making language models safer @AnthropicAI • prev CTO @WaveSenegal 🐧❤️Karina Nguyen @karinanguyen_
12K Followers 649 Following AI research & eng @AnthropicAI, prev. intern @nytimes, @square, @dropboxTim Dettmers @Tim_Dettmers
29K Followers 820 Following PhD Student at @UW. I blog about deep learning and PhD life at https://t.co/Y78KDJJFE7.10x’er @10x_er
28K Followers 1K Following ai, muay thai, running, lifting, climbing and still like 10 of your best guys at the officechris keefer @Dr_Keefer
24K Followers 3K Following ER Doc. Climate & anti-pollution activist. President Canadians for Nuclear Energy: https://t.co/HThtnXtbr6 Host Decouple Podcast: https://t.co/9xjJJaFA1fGeorge Hotz 🌑 @realGeorgeHotz
248K Followers 174 Following President @comma_ai. Founder @__tinygrad__kamilė @kamilelukosiute
220 Followers 88 Following where do we come from? what are we? where are we going?insane moments in bri.. @PoliticsMoments
134K Followers 1 Following UK politics isn't normal. Reminding people of the strange and downright insane moments that we have had to bear through the years. DM for ideasDominic Cummings @Dominic2306
295K Followers 2 Following peace abroad, regime change at home / maths circles / systems politicsMiranda Zhang @mirandahzhang
1K Followers 1K Following suffering reduction, AI safety, animal welfare, affordable housing. 💖 opinions my own.Joel Becker @joel_bkr
2K Followers 2K Following move fast and fix things. 'soccer'-me @MessiSeconds.isabelle 🪐 @isabelleboemeke
63K Followers 868 Following nuclear energy is clean energy 💎 in a simulation founder of @isodope @ZodiacMgmtAlexandr Wang @alexandr_wang
142K Followers 695 Following ceo at @scale_ai. rational in the fullness of timeToby Shevlane @tshevl
2K Followers 1K Following Research Scientist testing AI models for new capabilities at @GoogleDeepMind. Tweeting about AI and the future.Haydn Belfield @HaydnBelfield
4K Followers 2K Following @Cambridge_Uni researcher. Tweets about international security, AI governance, pandemics, nukes and climate change. @CSERCambridge & @LeverhulmeCFIAI Pub @ai__pub
72K Followers 343 Following AI papers and AI research explained, for technical people. Get hired by the best AI companies: https://t.co/MySVjUGOQ3Joscha Bach @Plinz
129K Followers 754 Following FOLLOWS YOU. Artificial Intelligence, Cognitive Architectures, Computation. The goal is integrity, not conformity. https://t.co/rFUNzdYXuKTristan Hume @trishume
6K Followers 330 Following Performance optimization lead @AnthropicAI. Profiling, distributed systems, dev tools, interpretability. [email protected]Tom Everitt @tom4everitt
1K Followers 596 Following AGI safety researcher at @GoogleDeepMind, leading the causal incentives group https://t.co/gBAjHPDr2jMatthew McDermott @MattBMcDermott
787 Followers 473 Following Berkowitz Postdoctoral Fellow at @HarvardDBMI; On the academic job market for Fall 2024Sebastian Farquhar @seb_far
2K Followers 140 Following Research Scientist @DeepMind - AI Alignment. Associate Member @OATML_Oxford and RainML @UniofOxford. All views my dog's.Zac Kenton @ZacKenton1
1K Followers 1K Following Research Scientist in AI safety at DeepMind. Views are my own and don't represent DeepMind.Nick @nickcammarata
60K Followers 734 Following interested in neural network interpretability and meditation𝐊𝐞𝐫𝐫𝐲 .. @KerryLVaughan
2K Followers 726 FollowingGrace Adams @gracevadams
1K Followers 282 Following Head of Marketing @GivingWhatWeCan Just doing my best 😌 All opinions are my ownNeel Nanda @NeelNanda5
13K Followers 89 Following Mechanistic Interpretability lead @DeepMind. Formerly @AnthropicAI, independent. In this to reduce AI X-risk. Neural networks can be understood, let's go do it!Peter Wildeford @peterwildeford
10K Followers 366 Following Pro forecaster w/ good track record. Seeking to understand + manage risks from advanced AI systems. - Co-CEO @RethinkPriors - Chief Advisory Executive @iapsAIChana @ChanaMessinger
4K Followers 2K Following aspiring to hawkish logic and quarrelsome empiricism the stakes and the world and the stars (I work at CEA but opinions here are my own)New Liberals 🌐🇺.. @CNLiberalism
84K Followers 1K Following Liberal, Open, and Radically Pragmatic. The home for the center-left. Part of @NewDemocracyJeremiah Johnson 🌐 @JeremiahDJohns
26K Followers 1 Following Internet analyst writing at Infinite Scroll. Founder @CNLiberalism. Carly Rae Jepsen stan. Send me the worst tweets on this site. https://t.co/xONcXY5PQvFrances Lorenz @frances__lorenz
4K Followers 538 Following ✨ I share my feelings, post lil jokes for the girlies, and often discuss effective altruism ✨ I also work on the EA Global team at CEA (views my own)Aella @Aella_Girl
205K Followers 369 Following ⚜️whorelord⚜️, vexworker, survey artist, way too earnest Discord: https://t.co/S1MaMdCwyKLawrence Atkins @latkins103
21 Followers 135 FollowingVaish @Vaishsg7
223 Followers 591 Following Personal opinions and off the cuff takes here. Substack: https://t.co/Sns3k71wpBQualy the lightbulb @QualyThe
7K Followers 319 Following Official Unofficial EA mascot. I'm here to make friends and maximise utility, and I'm all out of neglected altruistic opportunitiesNuño Sempere @NunoSempere
2K Followers 199 Following Researcher, forecaster, consultant, programmer. Disagreement is a ladder, measure is unceasing.eli ([email protected].. @erijohnt
171 Followers 271 Following hi i'm your sysadmin and i can be trusted with root access :) • he/they/she • touching computers @AnthropicAICaveat number 2, which is more recent, and (imho) clearer: alignmentforum.org/posts/TWorNr22…
Anthropic arguably initiated the growing consensus around non-publication of “capabilities” research—they publish openly, but only on safety/interpretability. Don’t lump them with FAIR 😜
9/ Already, many other LLM players like Adept, Character, and Cohere have not published the details of their models. Just blog posts. FAIR and Anthropic might remain as the only large open research labs.
Terrifying to realize our society's "world is about to end" alarm has been going off nonstop for our entire lives. Almost every psychologically healthy person has learned to tune it out. We burned through our x-risk coordination commons long ago.
This is a fruitful direction for alignment research. Great work by the @AnthropicAI team
It’s hard work to make evaluations for language models (LMs). We’ve developed an automated way to generate evaluations with LMs, significantly reducing the effort involved. We test LMs using >150 LM-written evaluations, uncovering novel LM behaviors. anthropic.com/model-written-…
2/ This is my first paper with Anthropic and I had a huge privilege to be working on this project w/ @EthanJPerez, @sam_ringer, @kamilelukosiute, and all other members! Some crucial charts from the paper:
We found a way to write language model (LM) evaluations w/ LMs. These evals uncover many worrying LM behaviors, some relevant to existential risks from AI. For example, LMs trained w/ RL from Human Feedback learn to state a desire to not be shut down. 🧵 x.com/anthropicai/st…
It’s hard work to make evaluations for language models (LMs). We’ve developed an automated way to generate evaluations with LMs, significantly reducing the effort involved. We test LMs using >150 LM-written evaluations, uncovering novel LM behaviors. anthropic.com/model-written-…
Use human preference data to fine-tune language models. It’s officially cool now! 🤠
NVIDIA is selling a new A800 to circumvent recent export restrictions. I think this might be a bad move, and the USG should probably extend the restrictions to cover a wider variety of GPUs with reduced bandwidth. Otherwise, China's AI progress might continue. Here's why🧵⬇️
Is it an EA retreat if you don't get asked at least once if what you're doing is net positive?
Our paper was accepted at #NeurIPS2022 🎉 We train a counterfactual model for predicting a single cell's response to drug molecules. chemCPA is accurate even for drugs that haven't been measured in a single-cell setting, and outperforms existing models 💊🧫
Happy to present our work on “Predicting Single-Cell Perturbation Responses for Unseen Drugs” at #ICLR2022 @MLDD_Workshop (spotlight) today at 12pm ET. The method, chemCPA, predicts phenotypic responses by including the molecular information about drugs:
I started finding the front page of the EA forum a little too busy to keep on top of everything that gets posted. In August there were over 450 forum posts! I made this app displaying every post with > 10 karma, so I can choose what to read. This happened overnight!
Content generation AI appears to be blowing robotics/control systems out of the water lately. Still no self driving cars, but there are going to be some extremely fast services in the near future
Re-up: “What’s the optimal number of bread loaves to ship to your town each year?” “That’s an absurd question—I have no idea. Haven’t you read Hayek? Let the market figure it out.” “OK. What’s the optimal number of foreign workers to immigrate to the US each year?” “140,000.”
The compute supply chain is one of the most interesting & important in the world. A thread🧵on @shin10173 & my new post on @Verfassungsblog: "Compute & Antitrust: Regulatory implications of the AI hardware supply chain, from chip design to cloud APIs" verfassungsblog.de/compute-and-an…
I'd probably be working on AI alignment even if I gave no weight whatsoever to the wellbeing of future people. Don't think the case for AI alignment work depends all that much on longtermism.
When people say that effective altruism focuses on weird stuff too much, I remember that it was people in and around EA who convinced me to worry about COVID back when all my real-life friends thought it was a silly, overblown concern or simply hadn't thought about it at all.
Shameless plug: I’m now ~2 months into my sabbatical-year visit to Anthropic, and I’m really impressed.
If nukes had been developed in 2022: 🚨 New paper alert! 🔥 Our BoOm BooM method 🎇 sets new record on the Nuke Map benchmark: 3.34M fatalities over London!💥🏆 Thanks to our generous uranium sponsors 💓☣️ ✅All diagrams & implementation details available for the community 🗒️📈
AGI was fringe in the ML world until last year when it went mainstream. I predict AGI safety will likewise become mainstream ML in 2-3 years