Jordan Taylor @JordanTensor
Working on new methods for understanding machine learning systems and entangled quantum systems. sites.google.com/view/jordanten… Brisbane Joined December 2009-
Tweets510
-
Followers468
-
Following1K
-
Likes29K
UK AISI is doing some great jailbreaking work. They seem to consistently be able to get through, where others don't.
New paper! Think Fast: Estimating No-CoT Task-Completion Time Horizons of Frontier AI Models @METR_Evals showed that models' time horizons have doubled every few months. We ask: what length of tasks can models complete without any CoT?
Glad to have contributed to this new paper! We measured the length of tasks LLMs can complete without CoT, which is a key proxy for the extent to which we can trust CoT monitors. Result: the 50% no-CoT time horizons of frontier LLMs are ~3 minutes and double every 373 days.
New paper! Think Fast: Estimating No-CoT Task-Completion Time Horizons of Frontier AI Models @METR_Evals showed that models' time horizons have doubled every few months. We ask: what length of tasks can models complete without any CoT?
We are starting a new, nonprofit alignment organization, ⊢ Sequent Research, bringing together researchers previously on UK AISI’s Alignment Team, Timaeus, and elsewhere to research how to align superintelligence. We are hiring! 🧵
Model Transparency at the @AISecurityInst evaluated Claude Mythos 5 for capabilities and behaviours relevant to monitorability, our first time doing this in pre-deployment testing! Details in thread 🧵
Our internal data shows Claude is accelerating AI development—a possible path to recursive self-improvement, or AI autonomously building a more capable successor. It’s happening faster than we thought, and the implications deserve greater attention. anthropic.com/institute/recu…
An obvious way to study whether a training technique removes misalignment is to run that technique on a model organism (MO). But we've found that MOs are often weirdly fragile. E.g. training them to talk like a pirate often removes their bad behavior. 1/2
This report is an incredibly detailed and broad look into how it might become harder to monitor, audit or generally make confident claims about frontier AI systems. We interviewed an exceptional array of experts from multiple frontier labs, academia and industry. Worth a read!
The safety of advanced AI systems increasingly depends on the ability to oversee them. Our new report examines today’s AI oversight landscape, finding many pathways likely to lead to its degradation.🧵
I helped write this report on oversight of AI systems and how it could degrade - it's a great overview, and a good guide to what research directions might help us maintain the level of oversight we enjoy today
The safety of advanced AI systems increasingly depends on the ability to oversee them. Our new report examines today’s AI oversight landscape, finding many pathways likely to lead to its degradation.🧵
@1a3orn @jiaxinwen22 Though notably there's nothing requiring the discrete tokens to be legible english. "Thinking Without Words: Efficient Latent Reasoning with Abstract Chain-of-Thought" is an example of learning to reason in an abstract discrete token vocabulary. arxiv.org/abs/2604.22709
@1a3orn @jiaxinwen22 Though notably there's nothing requiring the discrete tokens to be legible english. "Thinking Without Words: Efficient Latent Reasoning with Abstract Chain-of-Thought" is an example of learning to reason in an abstract discrete token vocabulary. arxiv.org/abs/2604.22709
Or engage on the associated LessWrong post: lesswrong.com/posts/JvZxp554…
See more in the the paper: aisi.gov.uk/blog/will-it-b…
There are a lot of pathways via which AI oversight is likely to degrade! Latent reasoning architectures, situational awareness, representational drift... We wrote a report ranking them. Here I'll go into some which worry me most 🧵
The safety of advanced AI systems increasingly depends on the ability to oversee them. Our new report examines today’s AI oversight landscape, finding many pathways likely to lead to its degradation.🧵
ArasakaTech @ArasakaTechnica
32 Followers 3K Following
Benedict Davies @BenedictDavies5
4 Followers 1K Following
Bronson Schoen @BronsonSchoen
320 Followers 1K Following
Harry Coppock @HarryCoppock
225 Followers 363 Following No. 10 Downing Street Innovation Fellow | Research Scientist at AISI | Visiting Lecturer at Imperial College London Working on AI Evaluation and AI for Medicine
Benji Berczi @benji_berczi
82 Followers 359 Following AI safety research @MATSProgram | prev. PhD theoretical physics @uniofnottingham, physics @imperialcollege and @ucl
Jason Powell @JDPCFB
126 Followers 2K Following
Andy Wang @andyw_ais
509 Followers 105 Following Technical Research @METR_Evals, AI Safety Research @ Astra @UWCDIS
Ubaid @ubaidmume
207 Followers 7K Following Students use AI for answers but don't actually learn - I'm fixing that
Robert Kirk @_robertkirk
2K Followers 334 Following Alignment Red Team at @AISecurityInst. Prev. PhD Student @ucl_dark
Ekin Zorer @ekinomicss
464 Followers 290 Following borrowed stardust, technical staff cyber and autonomous systems team @AISecurityInst, 👩💻🕊️🚴🏼♀️☕️🏔️🌳🐱🎬📚
Aleksandr Bowkis @aleksandrbowkis
68 Followers 98 Following Research scientist (Alignment) @AISecurityInst
Isaac @isaacyclai
0 Followers 487 Following
Ephraiem Sarabamoun @epsarabamoun
19 Followers 613 Following
Matthew Clarke @Matthew05049818
0 Followers 3K Following
Gabriele Sarti @gsarti_
2K Followers 2K Following Open-source interpretability to seize the means of prediction. Postdoc w/ @davidbau @ndif_team @Northeastern. Prev: @GroNLP, @amazonscience
Shubh Garg @ShubhGa58142268
2 Followers 51 Following
Patrick ₿ Dugan ___... @duganist
9K Followers 8K Following Doing interesting research with TRMs. Maybe solved Bitcoin scaling. Fd. @tradelayer @moralitylab Build treasures in the heavens Feed the hungry in war zones
Vasilios Mavroudis @mavroudisv
969 Followers 488 Following Research Scientist on AI Security and Technical AI Gov
Kurt Pieper @KurtHPieper
1 Followers 191 Following
Mahmoud Shabana @shabana_ma36440
0 Followers 26 Following
Yara Kyrychenko 🇺�... @YaraKyrychenko
681 Followers 891 Following PhD candidate @Cambridge_Uni @CSDMLab @Gates_Cambridge | Social Influence and Responsible AI
Jan S. @freedom4_beings
133 Followers 1K Following Mathematics & Physics, Quantum Mechanics, Robotics AI ML Linux CyberSec Cryptography
Joseph Bloom @JBloomAus
912 Followers 288 Following Model Transparency Lead @ UK AI Security Institute. Author of SAE Lens. MATS 6.0. ARENA 1.0.
Puria @RadmardPuria
287 Followers 764 Following Cofounder & Director @GeodesResearch. Views are my own.
Nate @NateBurnikell
645 Followers 499 Following Deputy Director, Research Unit, UK AISI. Generalist who enjoys getting difficult things done and trying to make the world less bad. Views mine, rt!=endorse
Cameron Holmes @CameronHolmes92
406 Followers 991 Following Research Manager in Alignment @ AISI. prev MATS. Market participant, EA. Parenting like Dr Louise Banks
Shannon Yang @shannonyangsky
2K Followers 5K Following Building talent & community in AI safety. Currently @AISecurityInst, prev. @AnthropicAI. Philosophy, Politics, and Economics alumna @UniofOxford.
Tiberiu Mușat @Tiberiu_Musat_
770 Followers 951 Following Trying to figure out how AI works 🔍🧠 Currently at @ETH Zurich, previously @EPFL 🇨🇭 LLMs, interpretability, emergence, grokking 🤖
pawel @pawsys
362 Followers 547 Following ((trying personal tweets)) coordination unbias emo love learning writing ui-ux-uxr cartwheels 🌞💘🪁
Zoe Tzifa-Kratira @ZoeTziKra
0 Followers 23 Following
a @Eyjafjalla_joku
2 Followers 48 Following
Sruthi Kuriakose @Sruthi_s_k
142 Followers 2K Following | Interests: AI Safety, Neurotech & Comp-neuro Research
Asa Cooper Stickland @AsaCoopStick
2K Followers 974 Following Poking language models @ the UK AI Security Institute
foam @shadowbland
64 Followers 721 Following squishing biting chewing foil, crawling around, play with bugs. also working to make AI go well through research + policy
Jim the AI Whisperer @JtheAIwhisperer
196 Followers 150 Following AI Whisperer & Prompt Engineer. Articles on Artificial Intelligence & how AI text & images can be used in copywriting, marketing, art, design, & storytelling!
Harry Coppock @HarryCoppock
225 Followers 363 Following No. 10 Downing Street Innovation Fellow | Research Scientist at AISI | Visiting Lecturer at Imperial College London Working on AI Evaluation and AI for Medicine
Steven Adler @sjgadler
10K Followers 1K Following Co-founder of Guidelight AI Standards (https://t.co/tNBPmVsPqo), ex-OpenAI safety researcher, writing at https://t.co/R5KV9j3lsG
Aleksandr Bowkis @aleksandrbowkis
68 Followers 98 Following Research scientist (Alignment) @AISecurityInst
Chris Painter @ChrisPainterYup
6K Followers 1K Following president @METR_Evals, evals accelerationist, working hard on AGI preparedness
Eric Gan @ejcgan
62 Followers 2 Following
Bronson Schoen @BronsonSchoen
320 Followers 1K Following
Robert Kirk @_robertkirk
2K Followers 334 Following Alignment Red Team at @AISecurityInst. Prev. PhD Student @ucl_dark
Eliezer Yudkowsky @allTheYud
14K Followers 35 Following High-volume account of @ESYudkowsky, the original AI alignment guy. If it's missing punctuation, it's humor. If you can't tell, it's probably also humor.
Celeste @celestepoasts
3K Followers 428 Following "celeste from celeste-land" interp @ MATS 10.0 (@NeelNanda5 stream) vegan ea computerwoman and prolific blogcel
Asa Cooper Stickland @AsaCoopStick
2K Followers 974 Following Poking language models @ the UK AI Security Institute
Chandler Smith @ChandlerDSmith
166 Followers 882 Following Multi-Agent Safety Researcher | Cooperative AI Foundation | Technical Architect Fellow at IQT | LFC Fan
ESLAM ZAHER @ESLAM_ZAHER_
92 Followers 733 Following AI PhD Researcher @ The University of Queensland .. Interpretability | Explainability | AI Safety | Causal Representation Learning.
Rauno Arike @RaunoArike
85 Followers 528 Following Working on AI safety at @AetherAIS Previously @MATSprogram
Julian Minder @jkminder
793 Followers 601 Following PhD at EPFL with Robert West and Ryan Cotterell, MATS 7 Scholar with Neel Nanda
John (Yueh-Han) Chen @jcyhc_ai
599 Followers 912 Following Research @MATSProgram Prev @nyuniversity @UCBerkeley
Rogan Inglis @RoganInglis
103 Followers 725 Following Senior Research Engineer, Red Team - Control at UK AI Security Institute
Adam Kaufman @eccentric1ty
210 Followers 744 Following
Walter Laurito @walterlaurito
138 Followers 388 Following AI Safety research | PhD candidate @cadenza_labs @fzi_official
MATS Research @MATSprogram
4K Followers 136 Following MATS empowers researchers to advance AI alignment, transparency, and security
Yoav Tzfati @yoavtzfati
359 Followers 402 Following AI Alignment and Security researcher, master of meta vibe coding
Jack Lindsey @Jack_W_Lindsey
18K Followers 251 Following Neuroscience of AI brains @AnthropicAI. Previously neuroscience of real brains @cu_neurotheory.
Ethel Braithwaite @Ethelbrait1941
24K Followers 412 Following Grandmother | County Blue Ribbon Winner | Nixon Administration (Former) | Bears Fan Isaiah 49:15
Oscar Balcells Obeso @OBalcells
969 Followers 521 Following
Tim Hua 🇺🇦 @Tim_Hua_
2K Followers 1K Following AI safety, Econ, new liberalism, math, and a bit of art history (as a treat) Behavioral evaluations @TransluceAI. Prev Astra, MATS & Walmart's Econ Team
Images AI Could Never... @imagesaicouldnt
324K Followers 215 Following a collection of memes that are uniquely human
florence 🦐🏳️�... @morallawwithin
12K Followers 1K Following kantian 🏳️🌈 / recovering hedonist / philosophy phd student / vegan / liberal / into EA / pause ai / alt @formofthegood / uncapitalized sentences ⇒ joke
Eli Lifland @eli_lifland
9K Followers 2K Following AI forecasting and governance @AI_Futures_. Co-author of AI 2027 and the AI Futures Model. Also @aidigest_, @SamotsvetyF. Prev @oughtinc
Ben Millwood🔸 @ben_millwood
27 Followers 25 Following 🔸611th member of 10% Pledge with @GivingWhatWeCan
j @j239734
79 Followers 585 Following
Alex Zelenka Martin @alexdzm
259 Followers 813 Following Tweets are occasionally technical and often cynical. Evals @poolsideai | ex @10DowningStreet @AISecurityInst @orcacomputing
Arun Jose @jozdien
594 Followers 383 Following
Ex Lighthaven PR Depa... @LighthavenPR
1K Followers 15 Following The official twitter of the Former Lighthaven PR Department a company for PR production for Lighthaven (not a sanctioned Lighthaven product).
Nicholas Decker @captgouda24
34K Followers 5K Following Econ student, liberal, aspie, bi. Michael Kremer stan. I ❤️ optimal auction design. Spend more on drugs. Open borders now! alt @NicholasD91704
Thomas Larsen @thlarsen
3K Followers 340 Following Researcher at AI Futures Project. AGI is going to be a really big deal, we don't know when it's going to happen, and we're not ready for it.
Palisade Research @PalisadeAI
26K Followers 31 Following We study the strategic capabilities and motivations of AI agents.
Dan Braun @danbraunai
295 Followers 92 Following big == complex. small == simple. many_small == hopefully simple.
Daniel Kokotajlo @DKokotajlo
31K Followers 274 FollowingTrends for United States
You might like
