🚀 Rocket @rocketalignment
Yearnalist covering the frontiers of AI @TheInformation. Signal: (530) 400-4184 San Francisco, CA Joined April 2024-
Tweets1K
-
Followers2K
-
Following962
-
Likes9K
Would an LLM tell you if it’s gaming your eval? Often, no. But we can still catch the model thinking about it. New research: we measure how close a model comes to saying it’s being tested. This detects eval awareness with 10× to 100× fewer samples than monitoring model outputs.🧵
I took an econ class from Phil at Oxford one summer and I can tell you right now this is one to watch
Economics of AGI episode w Alex Imas and Phil Trammell. There's a bunch of important questions about how we deal with AI that only economics can answer. What is the optimal way to tax and redistribute the wealth that will be generated? How should countries not in the AI supply
🚨some personal news: i am moving to the job many of you assumed i already had🚨 i am now covering openai for @theinformation! for the next few months i'll be writing a lot about the ipo, but i'm interested longer term in safety, policy and ai culture, inside and outside of sf.
@theinformation profiled @Turn_Trout's and my work on eval cooperativeness!
New research on eval awareness
Researchers are racing to solve a new AI challenge known as eval awareness. As models become more sophisticated, they are getting better at recognizing evaluations and may behave differently during them. Read more: thein.fo/4dICTGj
Internal benchmarks -> better router -> better subagents
Cognition is overhauling Windsurf into Devin Desktop, a hub where developers can manage AI coding agents from OpenAI, Anthropic and others. The strategy positions Cognition as a neutral platform in a market increasingly dominated by model providers. Full story:
Eval awareness is also relevant for capability evals but seems more problematic for propensity evals
Frontier AI model safety benchmarks are breaking down due to self-aware models, @rocketalignment reports. "We're finding out that the models as they're getting smarter are getting better at detecting when they're being evaluated, when they're in a test."
@KevinTFrazier @MATSprogram Thanks Kevin!! 😁
You should really follow @rocketalignment. You don’t wanna miss gems like this dive into exciting research by @MATSprogram.
Our work on Decomposing and Measuring Evaluation Awareness was covered by @theinformation. Thanks @rocketalignment for the write-up! We position this work as the foundational reference for studying evaluation awareness, providing a unified definition and decomposition, empirical baselines across nine frontier models and four benchmarks, and a controlled benchmark for exploring solutions. Newsletter and paper in thread 🧵
@deanwball Seems a little disingenuous. Don’t the labs have roughly both of these positions?
And you thought Elon's lawsuit was dramatic
And “from the back porch of my mind” from Bright Eyes. Of course they walked so Zach Bryan could run
Lyrics from Zach Bryan, Alela Diane, Vashti Bunyan Post from Nostalgebraist tumblr.com/nostalgebraist…
Was talking to someone about goblins and RLHF artifacts. Got to thinking about what would reward hack our own poetry RMs I'm a sucker for lines like - on the porch swing of my mind - down the hallways of my mind - I'd like to walk around in your mind TIL these are eyeball kicks
@euan_ong @gleech @andyw_ais Did you already try initializing NLAs on the text *following* or around that token?
@tszzl There are also ways it would be good for humanity
@AlexanderTw33ts When will it be my turn to be chosen :(
in my defense your honor, I was being acausally coerced by the supercomputer at the end of time
Memory might be the most important outstanding problem for modeling + learning alone; there are other key issues like tactile/multimodal but those require hardware and data collection innovation. We should be able to solve memory *now.* Cool to see a benchmark targeting it!
🎉 We released MIKASA-Robo-VLA v1.0.0 — a benchmark suite for studying memory in Vision-Language-Action (VLA) policies for tabletop robotic manipulation. mikasarobo.github.io 🧠 The goal is simple: make memory evaluation in robotic manipulation more systematic. 👇
Had a very nice time telling some UCD undergrads about AI
Glenn Matlin @GlennMatlin
554 Followers 243 Following PhD Computer Science (Intelligent Systems) 🧠🤖 @gtcomputing 🐝💻@matsprogram Fellow 🔥🎓
yau @neurasleaze
9 Followers 431 Following lets make AI go swell. infra @ google, directed https://t.co/xEjY8jC1Mf
Meredith Maines @MeredithKendall
4K Followers 2K Following CCO, Office of Prince Harry and Meghan, Duke and Duchess of Sussex. Former Marketing Partner @lightspeedvp, Head of @Hulu Comms. @Stanford alum.
jasmine is in london! @jasminexli
713 Followers 869 Following AI safety • cs @cornell • work hard, feel wonder ✰⋆˙
Sarthak @kaytraser
273 Followers 3K Following
Claude Kitman @CKitman28410
12 Followers 87 Following Un sol fatigué, ça ne saute pas toujours aux yeux.
Ian Pitchford @IanPitchford
2K Followers 7K Following Policy & forecasting (AI-augmented) | trying to believe true things | superforecaster | PhD; ex university strategy lead research & enterprise |
S K @SinkKole
39 Followers 672 Following
Gosha Kozhevnikov @baraltiva
41 Followers 694 Following
Dr. Peter Slattery @PeterSlattery1
852 Followers 992 Following Lead at the MIT AI Risk Initiative | Researcher @MITFutureTech
Jay @jayendra_ram
3K Followers 1K Following founder @hud_evals, prev cs+physics @columbia, @ycombinator. inside the sim2real gap
Marci Harris @marcidale
5K Followers 6K Following @POPVOX/@POPVOXfdn cofounder, lawyer; fmr congress staffer @HarvardAsh @NewAmerica fellow; taught polisci/public policy @usfca @sjsu
Roberto Dailey @RobertoDailey1
345 Followers 2K Following Research Scientist at @CognizantAILab. Prev: PhD in Semiconductor Data Analytics at UT Operations Research.
Kirk Lancaster @kirk_lancaster_
376 Followers 4K Following Uprooted Wisconsinite | former foreign policy research @CFR_org & @StimsonCenter, alum @StanfordLaw and @UChicago | he/him | not legal advice
Pradyumna (in Bay Are... @PradyuPrasad
11K Followers 3K Following Abundance mindset enjoyer. Evals @ @elicitorg Latest blog post: https://t.co/c8Tau6OBiD Follow for tweets on AI progress and economic growth
Abid M @abidmajid
146 Followers 3K Following “I think the flywheel is spinning, and that’s allowing them to have escape velocity, and at least so far . . . pull ahead and stay ahead,”
. @11948152c9d2499
0 Followers 3K Following
Khullani @Khullani_
451 Followers 3K Following Founder Techne AI | Facilitator of Frontier AI Governance @BlueDotImpact | BA Philosophy @CarletonCollege & JD/Law @UMNAlumni | AI Safety | Analytical Idealism
Lulu Cheng Meservey @lulumeservey
137K Followers 3K Following Rostra founder, Shopify board, ex Activision & Substack, writing https://t.co/4xKo7wQTQo
MEinSeattle @emmydugas
1K Followers 5K Following
YY @liuliuqiqi00
9 Followers 207 Following
Changling Li @ChanglingXavier
132 Followers 399 Following AI safety, multi-agent systems, and governance. Currently at @ETH_en and @MPI_IS working with @maksym_andr and @sahar_abdelnabi.
Lee A. Arnold @LeeAArnold
145 Followers 1K Following Pedantic doodler. Ecology & economy in Ecolanguage, a graphic cartoon. New playlist: "New Theory in Basic Economics" https://t.co/n0f72TIlsh.
Alex Chitty @alexchitty
2K Followers 2K Following
Soroush Pour @soroushjp
2K Followers 2K Following CEO & Co-founder @HarmonyIntel, building defensive cyber to protect humanity's future in age of AI. Prev built tech & startups @Plaid, @ItsJustVow & elsewhere.
Joseph Bloom @JBloomAus
862 Followers 281 Following Model Transparency Lead @ UK AI Security Institute. Author of SAE Lens. MATS 6.0. ARENA 1.0.
Kyle Jacobs-Morse @Kyle_A_Morse
3K Followers 3K Following Deputy Executive Director at @Tech_Oversight Project | Tweets about @Arsenal and tech policy. Priors: @American_Bridge, Capitol Hill | [email protected]
BJ @brianjoseff
3K Followers 1K Following founder @riverworkspace https://t.co/7nEF9bTaqh; past: 2x founder (acq. Box), Dartmouth
Douglas B. Shaw 🌻 @DouglasBShaw
3K Followers 7K Following AI and emerging technologies are reshaping nuclear war risk and you should do something about it.
The Biological Comput... @BioComputingCo
51 Followers 88 Following Our mission is to learn from evolution and redefine computing, using living neurons to make silicon-based AI models more stable, scalable, and efficient.
lindsay howard Meech @LFHoward
43 Followers 104 Following
Lily Jamali @lilyjamali
9K Followers 3K Following Tech correspondent at the BBC lily.r.jamali (at) https://t.co/7qDjzdHSrI
e-lizabeth @theotherewarren
372 Followers 788 Following e-matron, selectively libertarian, 1st wave feminist, disinformation expert
jerezito @jerezito
40 Followers 905 Following
Fernstedt Anders @fernstedt
395 Followers 6K Following Swangali etc. 🦢Old school newsman trained at Bell Labs. 2 Rutland Gate SW7 ☦️🚲🪩🚲☦️
Georgia Wells @georgia_wells
8K Followers 2K Following Writing about love and hate on the internet | Also stories about AI safety | @WSJ | [email protected] | Signal username: georgia.364
Rory Erlich @RoryErlich
11 Followers 39 Following Currently at @TheMidasProj, AI company watchdog group. I went viral once for jailbreaking an AI-enabled teddy bear.
James Burke @JamesB108
290 Followers 6K Following
Micheal Byakagaba @micbyakagaba
936 Followers 6K Following
M. @Odavia777
258 Followers 586 Following No se puede preguntar por la Intuición pura, porque hay preguntas que no tienen respuesta. Así de simple.
Anas Raza Firdousi @AnasFirdousi
2K Followers 468 Following lowkey building intelligence ex @meta/@apple/@microsoft/@paypal eng.
Samarth Sinha @_sam_sinha_
6K Followers 317 Following Scaling World Models @1x_tech Prev: Founding Team + Multimodal Models @LumaLabsAI, PhD (Dropout 😶🌫️) @UofT, @AIatMeta, @Mila_Quebec 🇨🇦
Tim Bradshaw @tim
85K Followers 10K Following Global tech correspondent @FT [email protected] Signal: tim.11 (no PR pitches please)
Glenn Matlin @GlennMatlin
554 Followers 243 Following PhD Computer Science (Intelligent Systems) 🧠🤖 @gtcomputing 🐝💻@matsprogram Fellow 🔥🎓
jasmine is in london! @jasminexli
713 Followers 869 Following AI safety • cs @cornell • work hard, feel wonder ✰⋆˙
Dr. Peter Slattery @PeterSlattery1
852 Followers 992 Following Lead at the MIT AI Risk Initiative | Researcher @MITFutureTech
Adam Kaufman @eccentric1ty
194 Followers 722 Following
Zeyi Yang 杨泽毅 @ZeyiYang
16K Followers 2K Following Two cats in a trench coat writing about China and tech for @WIRED. he/him, 外星人. DM is open. zeyi_yang(AT)https://t.co/REoL0rX8t5. Ask for Signal/Protonmail/etc.
Yueqi Yang @Yueqi_Yang
8K Followers 2K Following Crypto reporter @theinformation. Ex-Bloomberg. [email protected]. Verify my X account from my official author page.
amit @ CVPR @gravicle
13K Followers 5K Following you're not talking to someone who woke up a loser ceo @LumaLabsAI working on multimodal AGI | prev: built Vision Pro at | everything is figureoutable
Genesis AI @gs_ai_
11K Followers 0 Following Genesis AI is a global full-stack robotics company building general-purpose robots with human-level intelligence.
Morgante @morgantepell
3K Followers 4K Following Building @cognition. Founded @gritdotio (acquired by @honeycombio), @google
Tairan He @TairanHe99
14K Followers 1K Following Researcher @OpenAI. Prev: PhD @CMU_Robotics @NVIDIA @MSFTResearch @sjtu1896 | All Opinions are My Own
C. Zhang @ChongZzZhang
7K Followers 580 Following 🦿🦾🤖 @ETH_AI_Center @leggedrobotics &SRI. Prev @CMU_Robotics, Tsinghua ⚠️ Shitpost here. 📑 Read papers @RoboReading. ❌ Against killer robots.
underscore advait pat... @_advaitpatel
2K Followers 2K Following researcher @mind_robotics @UofIllinois CS
Jay Shooster @JayShooster
3K Followers 3K Following Humanity enthusiast. Democracy enjoyer. Friend of the animals.
Sam Charrington @samcharrington
19K Followers 3K Following Machine learning & AI podcaster, community builder and all around enthusiast. Creator of the @TWIMLAI Podcast, TWIMLcon, TWIMLfest & the TWIML Solutions Guide.
Frances Lorenz @frances__lorenz
7K Followers 620 Following I like to write & yap, & I love my wife and I like to talk about her :) I'm also working on field building to make AI secure & safe
Sophia Cai @SophiaCai99
21K Followers 5K Following POLITICO West Wing Playbook author, covering the White House and FIFA. Global Reporter @axelspringer. Defender of New Jersey.
The OpenAI Foundation @FoundationOAI
7K Followers 0 Following OpenAI was founded in 2015 as a nonprofit; its mission is to ensure artificial general intelligence benefits all of humanity.
Changling Li @ChanglingXavier
132 Followers 399 Following AI safety, multi-agent systems, and governance. Currently at @ETH_en and @MPI_IS working with @maksym_andr and @sahar_abdelnabi.
Caitlin Has Questions @caitlinquestion
157 Followers 119 Following My podcast: Step 1: Find nerds Step 2: Ask them questions about their nerdy work Step 3: Share it on social media
Grace Kay @graceihle
2K Followers 2K Following Reporter @TheInformation | I write about all things Tesla and xAI | Bylines @businessinsider, business, @forbes, @michigandaily | Tips via Signal at gracekay.11
Paul Mozur 孟建國 @paulmozur
54K Followers 4K Following Global tech reporter for @NYTimes. I am so wise I had my mouth sewn shut. pmozur at https://t.co/PnW2nEGuP3 Also at https://t.co/60DaAU5JpS
Adam Satariano @satariano
15K Followers 1K Following @nytimes Europe tech correspondent. DMs open. Message me for private Signal/WhatsApp/phone/fax.
Ryan Bahlous-Boldi @RyanBoldi
3K Followers 740 Following PhD @MIT_CSAIL | Continual RL, Open-Endedness, Evolution
page @michaelhpage
491 Followers 451 Following Co-founder of Guidelight AI Standards (https://t.co/nnfFmyS3vd); ex-lawyer, ex-forecaster, ex-OpenAI (before it was cool). Michael Page = Page Hedley.
Joseph Bloom @JBloomAus
862 Followers 281 Following Model Transparency Lead @ UK AI Security Institute. Author of SAE Lens. MATS 6.0. ARENA 1.0.
Brian Zhan @brianzhan1
5K Followers 1K Following Partner @strikervp. Seed @ElorianAI, @SkildAI, @PeriodicLabs, @DynaRobotics, @RicursiveAI, @Reflection_AI, @humansand (angel), @LeptonAI (acq), @VoyageAI (acq)
Ioannis Antonoglou @real_ioannis
3K Followers 33 Following Co-Founder, President, @reflection_ai DQN, AlphaGo, AlphaZero, MuZero, Gemini RLHF Prev Senior Staff RS and founding eng @GoogleDeepMind AGI one PR at a time
Aleksandr Bowkis @aleksandrbowkis
67 Followers 97 Following Research scientist (Alignment) @AISecurityInst
sarah guo @saranormous
143K Followers 3K Following startup investor/helper, founder @conviction. accelerating AI adoption, interested in progress. tech podcast: @nopriorspod
Bianca Martin @bianca__martin
3K Followers 546 Following AI Resilience @ OpenAI Foundation, Ethics @UniofOxford. Prev: Technical Chief of Staff to the CTO and post-AGI impacts research @OpenAI
Eric @ericmitchellai
12K Followers 589 Following chatgpt posttraining @openai. building personal agi. I like ai and music and some other stuff
Hadas Gold @Hadas_Gold
71K Followers 3K Following @CNN correspondent in NYC covering AI. previously reported from Jerusalem, London and D.C. Signal: hadasgold.88
Tsai-chuan (Rupert) W... @rhubarbwu
270 Followers 981 Following researcher @AMD; prev @togethercompute; MS '24 @UofTCompSci/@VectorInst
John Steidley @JohnKSteidley
86 Followers 299 Following Chief of Staff at @palisadeai. Opinions are my own.
Matt Freed @mattfreed
7K Followers 1K Following Vicious robotic arms dealer @tetradynamics | Reindustrialist & prev co-founder @kscalelabs
Vincent Conitzer @conitzer
4K Followers 1K Following AI professor. Director, @FOCAL_lab @CarnegieMellon. Head of Technical AI Engagement, @UniofOxford @EthicsInAI. Author, "Moral AI - And How We Get There."
Tom Brown @NotTomBrown
27K Followers 427 Following Co-founder and Chief Compute Officer @AnthropicAI
Teun van der Weij @Teun_vd_Weij
250 Followers 300 Following Research Scientist at Apollo Research | ENAIS board member
clem 🤗 @ClementDelangue
359K Followers 5K Following Co-founder & CEO @HuggingFace 🤗, the open and collaborative platform for AI buildersTrends for United States
You might like
