AI policy is a bit like proxy warfare for wider political goals. Every socioeconomic issue finds a new home as AI policy, every concern fits under inflated definitions, and any risk justifies regulation. I think the best antidote to this remains considering everything issue per…
New @GoogleDeepMind MechInterp work! We introduce Gated SAEs, a Pareto improvement over existing sparse autoencoders.
They find equally good reconstructions with around half as many firing features, while maintaining interpretability (CI 0-13% improvement). Joint w/ @ArthurConmy
Personally I would be happy with humanity getting 1% of the reachable universe and AIs getting the rest. I just don’t know what safeguards could ensure that we *keep* that 1%.
Today we're launching PRISM, a new resource to diversify the voices contributing to alignment. We asked 1500 people around the world for their stated preferences over LLM behaviours, then we observed their contextual preferences in 8000 convos with 21 LLMs arxiv.org/abs/2404.16019
Excited to be offering @ESRC grants of £280k to £800k for work using AI-powered research tools to address major social or economic issues.
The funding can be used for research, methodological development, and research infrastructure.
Find out more here:
ukri.org/opportunity/es…
Excited to share our work at @GoogleDeepMind!
We propose Naturalized Execution Tuning (NExT), a self-training method that drastically improves the LLM's ability to reason about code execution, by learning to inspect execution traces and generate chain-of-thought rationales 🧵👇
Disastrous news from the Philippines- the Court of Appeals has sided with Greenpeace and the antis in blocking vit-A enhanced Golden Rice, which is now in widespread cultivation in the country. All planting must now stop, harming the interests of millions.
philrice.gov.ph/philrice-ackno…
‘A European security official who regularly works with the BfV said the sharing of classified information between services on the continent was now “absolutely vital to almost every successful action that has been taken [against Russia and China]”.’
ft.com/content/4ec876…
LLMs can hallucinate and lie. They can be jailbroken by weird suffixes. They memorize training data and exhibit biases.
🧠 We shed light on all of these phenomena with a new approach to AI transparency. 🧵
Website: ai-transparency.org
Paper: arxiv.org/abs/2310.01405
here is an interface for generating chat interfaces with arbitrary personas. you can also initiate conversations from logs (format doesnt matter bc agi), so you can e.g. import claude chats into websim
websim.ai/c/oFskF68gjd7n…
On Feb 15, 2013, an eerie coincidence occurred: at the same time that a 45m asteroid narrowly missed Earth, a 20m meteor exploded over Russia with the force of 30 Hiroshima bombs, injuring 1000.
What if AI could help us build a more cooperative society? 🌱 Our latest research reveals how a deep-RL agent incentivizes humans to behave sustainably in a common-pool resource game: arxiv.org/abs/2404.15059.
Check it out! #AI#sustainability
The CMA is going to be doing a lot more of this once the government finishes passing the Digital Markets Unit legislation. I'm sceptical that a competition regulator getting involved will speed up progress in AI or other advanced technologies, but that's a minority view.
The CMA is going to be doing a lot more of this once the government finishes passing the Digital Markets Unit legislation. I'm sceptical that a competition regulator getting involved will speed up progress in AI or other advanced technologies, but that's a minority view.
Introducing “FlowMap”, the first self-supervised, differentiable structure-from-motion method that is competitive with conventional SfM like Colmap!
cameronosmith.github.io/flowmap/
IMO this solves a major missing piece for internet-scale training of 3D Deep Learning methods.
1/n
4K Followers 2K Following@Cambridge_Uni researcher. Tweets about international security, AI governance, pandemics, nukes and climate change. @CSERCambridge & @LeverhulmeCFI
4K Followers 537 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)
145K Followers 7K Following“A beacon of clarity”. Spoke at US Senate AI Oversight committee. Founder/CEO Geometric Intelligence (acq. by Uber). Rebooting AI & Taming Silicon Valley.
87K Followers 2K FollowingAsia business & finance editor @TheEconomist. Co-host of our Money Talks podcast. Just a lad from Leeds with a lust for markets, economics and history.
2K Followers 1K FollowingDirector of https://t.co/gCEDoKdKBT at Uni of Cambridge | Researching Big Risks, and impacts of AI & emerging tech. Opinions own
355 Followers 3K FollowingResearch Officer at Faculty of Economics, The University of Tokyo. Keywords: Entrepreneur/OpenAI/Quantum/Crypto/Analytics/Consulting. Views are my own.
1K Followers 487 FollowingSenior Researcher in @InstituteForGov’s civil service team. Las Vegas Sphere enthusiast. Likes are sometimes bookmarks. All views my own.
111 Followers 948 FollowingPhysics and abstract nonsense. I eat the complex systems on my pizza. Emergentist. Music and cooking nerd. Penguins can delay the past. — Exploring the void.
321 Followers 879 FollowingResearcher at Microsoft Africa - working on task alignment and evaluation of LLMs | Ex- @benevolent_ai | Ex- NEC Labs Europe | Ex-@UCL | @so_innovate
10K Followers 7K FollowingResearch Scientist @Google working on Gemini AI & multimodal LLM applications/ PhD @UW / former resident in AI / ML @msft @google @Theteamatx, alumn @mit
877 Followers 4K FollowingCS Engineer @PunePict
studying MSc @thehertieschool
Working student in #legaltech
Member @governancepost
Data| Tech| Ethics| Public Policy
349 Followers 459 FollowingProject Director @UCSD, working with teams @JPAL_SA, managing @riseprogramme India CRT | PhD @HarvardSoc | data for better policies | read-travel-meet-learn |
206 Followers 8K FollowingMehr Licht !
_『Macht doch den Fensterladen im Schlafgemach auf, damit mehr Licht hereinkomme.』(Goethes Tod)
_KAWAII is justice.Thank you so much!
4K Followers 1K FollowingResearcher @DeepMind working on safer Conversational AI | Honorary Professor @HeriotWattUni | Co-founder @helloalana | mother of dragons | own opinions only
912 Followers 782 FollowingSenior Research Scientist @BytedanceTalk working on Responsible AI prev. @oxcsml @UnioxOxford, @amazon, @apple, @bloomberg
All opinions are my own
175K Followers 89 FollowingThe original AI alignment person. Missing punctuation at the end of a sentence means it's humor. If you're not sure, it's also very likely humor.
4K Followers 2K Following@Cambridge_Uni researcher. Tweets about international security, AI governance, pandemics, nukes and climate change. @CSERCambridge & @LeverhulmeCFI
4K Followers 537 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)
145K Followers 7K Following“A beacon of clarity”. Spoke at US Senate AI Oversight committee. Founder/CEO Geometric Intelligence (acq. by Uber). Rebooting AI & Taming Silicon Valley.
7K Followers 319 FollowingOfficial Unofficial EA mascot. I'm here to make friends and maximise utility, and I'm all out of neglected altruistic opportunities
3K Followers 5K FollowingLabour Peer, social entrepreneur, community leader, social justice defender, publisher, educationalist, and plane spotter. All views my own
4K Followers 1K FollowingResearcher @DeepMind working on safer Conversational AI | Honorary Professor @HeriotWattUni | Co-founder @helloalana | mother of dragons | own opinions only
3K Followers 3K FollowingWe are a non-profit organization dedicated to helping governments and citizens make better decisions through open platforms and artificial intelligence.
54 Followers 25 FollowingAmericans for Responsible Innovation (ARI) is a nonprofit organization dedicated to AI policy advocacy in the public interest.
2K Followers 2K FollowingFounder and CEO of the Inclusive Abundance Initiative: https://t.co/2N11JTBvya,#HistoryOfScience,❤️🏃🏻♂️🚴🏻♂️🥾,26.2 in 2:59:29,3 boys w/@leorahorwitzmd
5K Followers 3K FollowingText mode, pixel art, games and tools
He/him
Support me on https://t.co/drc9Er6WQb
https://t.co/pGTruQSMFD
https://t.co/syNf3pNqNH
⏚
552 Followers 976 FollowingTech and society outside the global north. Senior research scientist at DeepMind. Past: Meta, Princeton, TVR2C at WPRB, Peace Corps, USDOJ
3K Followers 848 FollowingEntangled Others is an entanglement w/ the more-than-human, where diversity & inter-connectedness are nurtured & engaged through generative art.
186 Followers 173 FollowingResearch @ UK AISI. AI/ML DPhil @ Oxford. Former eng+DS. Looking to make a positive, lasting difference to the world. I like pizza, friends, symphonies.
50K Followers 2K FollowingResearch Scientist at Google DeepMind. I tweet about deep learning (research + software), music, generative models (personal account).
23K Followers 11K FollowingDissecting beatmaking, DJing, music production, rapping, and sampling. Articles, playlists, and Twitter threads. Curated/written by @ginosorcinelli.
3K Followers 0 FollowingGo beyond your mind's limitations. Personalized AI powered by what you've seen, said, and heard. Founders: @dsiroker, @brettbejcek, and @stammy