Caleb Withers @CalebWithersDC
AI safety & stability/tech & natsec @CNASdc. @GeorgetownCSS alum. Views my own. cnas.org/people/caleb-w… Washington, DC Joined February 2023-
Tweets721
-
Followers416
-
Following615
-
Likes12K
I'm delighted to present my NEW @CNASdc report: Beyond China's Black Box: Five Trends Shaping Beijing’s Foreign and Security Policy Decision-Making Under Xi Jinping Read it here: cnas.org/publications/r…
Huge activity on the national security front in Congress. Comments on Ukraine, Israel, Tik Tok & more from the team @CNASdc. Highlights to follow....cnas.org/press/press-no…
Very excited about a new paper from @jordanschnyc and @arrian_ebrahimi on the R&D agenda for CHIPS The National Semiconductor Technology Center has received a lot less attention than the big manufacturing subsidies, but it has the potential to be the most impactful part of CHIPS
Zuck on: - Llama 3 - open sourcing towards AGI - custom silicon, synthetic data, & energy constraints on scaling - Caeser Augustus, intelligence explosion, bioweapons, $10b models, & much more Enjoy! Links below
My coverage of the @dwarkesh_sp interview with Zuckerberg, and all things Llama 3. shorturl.at/fxMY5
Read @CalebWithersDC commentary on the implications of “open-weight” AI models, how open-weight models realize some of the benefits of open-source software, and important distinctions around the potential misuse and opacity of large AI models. More: cnas.org/publications/c…
.@Microsoft's investment in @G42ai will enhance the UAE’s position as a leading #AI hub and provide digital infrastructure and services to help underserved nations innovate and grow. We will combine world-class technology with world-leading standards for safe, trusted, and…
SCOOP: A TikTok data scientist says he was assigned a Seattle-based manager on paper, while actually reporting to a Beijing-based ByteDance executive, who ordered him to regularly email U.S. data in spreadsheets to ByteDance workers in China during 2022. fortune.com/2024/04/15/tik…
There's been a lot of buzz about "emergent abilities" in large language models, including some media exaggeration. I took a crack at explaining the different perspectives. 🧵 cset.georgetown.edu/article/emerge…
I’m super excited to release our 100+ page collaborative agenda - led by @usmananwar391 - on “Foundational Challenges In Assuring Alignment and Safety of LLMs” alongside 35+ co-authors from NLP, ML, and AI Safety communities! Some highlights below...
Just released! The #AIIndex2023 rounds up the latest trends in AI. This year’s report introduces more original data than any previous edition, a new chapter on AI public opinion, and more. Here are the top takeaways: ↘️ stanford.io/3m3gStO
This Google Scholar plugin is meaningfully accelerating my paper reviews. Clicking on a citation immediately pops up a helpful info box from Scholar, no more going back and forth to the references section, googling a title, reading the abstract, looping back, etc.
This Google Scholar plugin is meaningfully accelerating my paper reviews. Clicking on a citation immediately pops up a helpful info box from Scholar, no more going back and forth to the references section, googling a title, reading the abstract, looping back, etc. https://t.co/jVwKRDOQ9K
"I detect an undercurrent of self-doubt among some Americans about what your role in the world should be...At a time when our world is at history's turning point...Freedom and democracy are currently under threat around the globe." — Japanese PM Fumio Kishida addresses Congress
So good to see Google start to offer Confidential Computing VMs for AI. Data/code integrity, confidentiality & remote attestation are going to be super useful for the AI workloads of the future.
So good to see Google start to offer Confidential Computing VMs for AI. Data/code integrity, confidentiality & remote attestation are going to be super useful for the AI workloads of the future.
GPQA evaluation of new GPT-4 Turbo release!
GPQA evaluation of new GPT-4 Turbo release!
This is an important point by @carlbfrey and @maosbot :
This is an important point by @carlbfrey and @maosbot : https://t.co/URJVhW6JSS
Introducing Udio, an app for music creation and sharing that allows you to generate amazing music in your favorite styles with intuitive and powerful text-prompting. 1/11
What should we be evaluating besides model capabilities? What about: how well we can predict what models will do in the real world? I explore this question in a new @GovAI_ blog post 🧵
Great write-up from @chrisrohlf re: on-chip governance. Agreed with almost everything, except: > the implementation must be near perfect to provide long term value > must be extremely costly to defeat This will be true in some cases, but depends on use case.
Great write-up from @chrisrohlf re: on-chip governance. Agreed with almost everything, except: > the implementation must be near perfect to provide long term value > must be extremely costly to defeat This will be true in some cases, but depends on use case.
I give (hopefully useful?) advice to folks on careers in AI policy a lot, and many themes come up repeatedly. So I wrote this post which includes FAQs as well as more general advice. Hope some folks find it helpful! medium.com/@miles_24227/f…
XeniaCarter @Ym48vSsmD4h3S
2 Followers 191 Following
Sloughey @SlougheyhDw
0 Followers 127 Following
Ceethosh @ceethosh13043
1 Followers 176 Following
Ophelia Thruthley @thruthley50023
82 Followers 5K Following
DrucillaWaszak @DrucillaWa43322
14 Followers 776 Following
Augustine Steveson @AugustSteves
29 Followers 5K Following
Norma Stigers @stige_nor
71 Followers 5K Following
Jindong Gu @Jindong73504766
296 Followers 891 Following Senior Researcher @UniofOxford Faculty Researcher @GoogleResearch #ResponsibleAI #AISafety #GenAI Homepage: https://t.co/YOSVO3jb6h
Jewel Shiring @je_shirin
89 Followers 5K Following
kazoomarc @datamoree
626 Followers 5K Following • random-digital-art-ish — FB: https://t.co/ZEI4oo3qwq — IG: https://t.co/Rh4056hSDd
Stefania Hackler @HackleStefani
80 Followers 5K Following
Matt Reardon @Mjreard
354 Followers 207 Following Advising team @80000hours. Host of the Actually After Hours Podcast. Will go back to having cool bios when I'm popular.
Audrey Fyles @FylesAud
57 Followers 5K Following
Ivy-mae Thronson @ThronIvy
83 Followers 5K Following
Giselle Goudie @GiselGoudi
59 Followers 5K Following
Muriel @wakaida62949660
3 Followers 752 Following
Ava-rose Aikens @AvaAikens15100
51 Followers 5K Following
Russ @RussellWeigley
6 Followers 1K Following
Daniel Cebul @DCebul
404 Followers 1K Following Political Science PhD student @MIT, previously @GeorgetownCSS and @KenyonCollege. @Browns, @Cavs, and @ManUtd fan
CipherLaw @CipherLaw
709 Followers 3K Following AI + Open Source + Cybersecurity intellectual property and techlaw. Counsel to disruptive technology companies in fast-paced, dynamic and chaotic environments.
Alivia Mardirossian @AliviaMard39769
33 Followers 5K Following
Madeline Boeck @boe_madeli
47 Followers 5K Following
Connor Dunlop @cp_dunlop
596 Followers 2K Following EU Policy Lead @adalovelaceinst. Views are my own!
This American Adam @adamconner
16K Followers 9K Following VP Tech Policy at @amprog @CAPaction. Before @HarvardIOP, @SlackHQ, Brigade, @Facebook, campaigns. Always and forever a New Mexican. Zia’s dad. Opinions my own.
Violet Leger @VioletLege86784
98 Followers 5K Following
Fiadh Petronio @FiaFiadh
53 Followers 5K Following
Timothy B. Lee @binarybits
43K Followers 1K Following Reporting on AI and the future of the economy. Computer science masters degree from Princeton. @arstechnica alum. Subscribe to my AI newsletter!
Andrew Trask @iamtrask
74K Followers 190 Following @openminedorg, @GoogleDeepMind ethics team, @OxfordUni phd candidate, @UN pet lab, @GovAI_, creator of #GrokkingDeepLearning, NALU, and sense2vec
Joshua Saxe @joshua_saxe
3K Followers 982 Following AI+cybersecurity at Meta; past lives in academic history, labor / community organizing, classical/jazz piano, hacking scene
Isha Gupta @IshaLGupta
1K Followers 2K Following Research Analyst, @USIP South Asia (she/her) Previously: @StimsonCenter, @SAVoices, @Wellesley | views mine/do not represent my employer, RT ≠ E, etc.
Lulu Shamp @LuluShamp4248
73 Followers 5K Following
Gavin Brown @gavinobrown
539 Followers 717 Following Director of Scholastic's Home Base. Author of Monster Club: Hunters for Hire and Josh Baxter Levels Up. Creator of Blindscape mobile game. He/Him.
Davis Brown @davisbrownr
352 Followers 980 Following Research in interpretability, science of deep learning, safety and security @pnnlab. Opinions my own.
Noelia Petrea @NoeliaPetr68643
81 Followers 5K Following
Aris Richardson @richardson_aris
38 Followers 54 Following AI policy researcher. Winter Fellow @GovAI_ ‘24 | @ucberkeley ‘22
Jayde Bresciani @BrescianiJ87279
84 Followers 5K Following
Jina Lee @JinaLeeSoc
1K Followers 2K Following Ph.D. @ua_sociology | Incoming Asst. Prof @UofIllinois Sociology | Computational Social Science, Sociology of Science, Gender, Medical Sociology | She/Her/Hers
Willetta Lesslie @LesslWillett
55 Followers 5K Following
Eva @MarinaThac78783
14 Followers 202 Following For things we can control, we should remain cautious, and for things we cannot control, we should remain optimistic.
Ena Maclauchlan @maclauchla_e
39 Followers 5K Following
amit PRIVATE 𝕏 @amitisinvestPRI
13 Followers 1K Following only private elite here‼️I will mainly talk about stock market, price action analysis etc. ONLY FOR A FEW PEOPLE. MAIN PAGE:@amitisinvesting
Cadi Vrabel @cad_vrab
84 Followers 5K Following
Jindong Gu @Jindong73504766
296 Followers 891 Following Senior Researcher @UniofOxford Faculty Researcher @GoogleResearch #ResponsibleAI #AISafety #GenAI Homepage: https://t.co/YOSVO3jb6h
Grace Kind @kindgracekind
2K Followers 2K Following AI navel-gazer / Ideonomy evangelist / navigator of uncertain waters
palcu @AlexPalcuie
1K Followers 422 Following By day, I'm a software engineer/SRE making sure Google Cloud stays afloat. By night, I'm shitposting or leetcoding. Big Matt Levine and Scott Alexander fan.
Matt Reardon @Mjreard
354 Followers 207 Following Advising team @80000hours. Host of the Actually After Hours Podcast. Will go back to having cool bios when I'm popular.
U.S.-China Commission @USCC_GOV
21K Followers 542 Following USCC is a bipartisan Commission created by Congress to monitor & report on trade, economic, defense, and foreign affairs issues in the U.S.-China relationship.
Jacob Helberg @jacobhelberg
23K Followers 2K Following commissioner @uscc_gov • senior policy advisor to @PalantirTech CEO • author of The Wires of War (@simonschuster) • ex @Google (views are my own)
Dean Woodley Ball @deanwball
452 Followers 1K Following AI Research Fellow @mercatus, author of Hyperdimensional (see link) formerly @HooverInst @CoolidgeFdn @SmithSoc @ManhattanInst Thoughts my own, like != endorse
Limitless @LimitlessAI
3K Followers 0 Following Go beyond your mind's limitations. Personalized AI powered by what you've seen, said, and heard. Founders: @dsiroker, @brettbejcek, and @stammy
Dan Siroker @dsiroker
33K Followers 3K Following Co-Founder & CEO of @LimitlessAI: a personalized AI powered by what you’ve seen, said, or heard. Co-Founder of @Optimizely. Three kids under 6. Amazing wife.
Mike Gallagher @RepGallagher
53K Followers 457 Following
Elbridge Colby @ElbridgeColby
62K Followers 4K Following Principal, Marathon Initiative. Former Pentagon, 2018 National Defense Strategy, inter alia. Author of Strategy of Denial. Views my own.
Naman Jain @StringChaos
904 Followers 906 Following CS PhD @UCBerkeley | Projects - R2E, LiveCodeBench, Chatbot-Arena Coding, RAFT, Data Quality | Past: @AWS @MSFTResearch @iitbombay
Connor Dunlop @cp_dunlop
596 Followers 2K Following EU Policy Lead @adalovelaceinst. Views are my own!
Deric Cheng @0xDeric
501 Followers 191 Following Leading the Governance Research team @CvgAnalysis Previously: | 5th engineer @AlchemyPlatform | Founded @web3university @solaurafest | Researcher @google
Convergence Analysis @CvgAnalysis
15 Followers 38 Following Designing a safe and flourishing future for humanity in a world with transformative AI
Tolga Bilge @TolgaBilge_
2K Followers 571 Following Leader of https://t.co/IiJWWuiWhD & https://t.co/Ep1VpsILyj — @Superforecaster, @SamotsvetyF, @Swift_Centre, @INFERpub — Mathematics graduate of @UiB & @UnivofStAndrews
Phil Venables @philvenables
12K Followers 694 Following Tweets about cybersecurity, resilience & enterprise risk - at scale. CISO - Google Cloud + 3 x CISO (25 yrs), Board Director, Chief Risk Officer Tweets=own.
Andrew Trask @iamtrask
74K Followers 190 Following @openminedorg, @GoogleDeepMind ethics team, @OxfordUni phd candidate, @UN pet lab, @GovAI_, creator of #GrokkingDeepLearning, NALU, and sense2vec
Neil Chilson ⤴️�.. @neil_chilson
7K Followers 2K Following Lawyer, computer scientist & author of book 'Getting Out of Control.' Was chief technologist @FTC, now Head of AI Policy at Abundance Institute.
udio @udiomusic
28K Followers 0 Following
Joshua Saxe @joshua_saxe
3K Followers 982 Following AI+cybersecurity at Meta; past lives in academic history, labor / community organizing, classical/jazz piano, hacking scene
Center for AI Policy @aipolicyus
367 Followers 2 Following A nonpartisan research organization dedicated to mitigating the catastrophic risks of AI through policy development and advocacy.
Davis Brown @davisbrownr
352 Followers 980 Following Research in interpretability, science of deep learning, safety and security @pnnlab. Opinions my own.
Yi Tay @YiTayML
29K Followers 97 Following chief scientist / cofounder @RekaAILabs 🫠 past: research scientist @google brain 🤯 currently learning to be a dad 🍼
Ethan Caballero is bu.. @ethanCaballero
8K Followers 2K Following ML PhD student @Mila_Quebec ; previously @GoogleDeepMind
Aris Richardson @richardson_aris
38 Followers 54 Following AI policy researcher. Winter Fellow @GovAI_ ‘24 | @ucberkeley ‘22
Frontier Model Forum @fmf_org
123 Followers 5 Following A coalition of industry leaders, committed to the safe and responsible development of frontier AI
david rein @idavidrein
2K Followers 985 Following Sentio ergo sum. AI alignment research at NYU, early employee @cohere
Tom Shugart @tshugart3
30K Followers 725 Following Defense analyst, former submariner, pilot. @cnasdc Adjunct Senior Fellow. Founder, Archer Strategic Consulting. Opinions my own. RT=/=endorsement.
Joseph Cox @josephfcox
88K Followers 3K Following Hacking/crime/privacy journalist. Author of forthcoming DARK WIRE. Co-founder of @404mediaco. Signal: +44 20 8133 5190. Email: [email protected]
Santi Ruiz @rSanti97
7K Followers 4K Following He was so joly of his joyfnes | editing @ifp, writing https://t.co/ZiChKOFfhH and https://t.co/ucgoUX3exu | dms open | ☧
Marci Harris @marcidale
5K Followers 6K Following @POPVOX/@POPVOXfdn cofounder, lawyer; fmr congress staffer @HarvardAsh @NewAmerica fellow; adjunct @usfca lecturer @sjsu
Saad Siddiqui @Saad97Siddiqui
173 Followers 712 Following Researcher, Safe AI Forum | Sino-Western AI Safety and Governance Past: @Bainandcompany, @schwarzmanorg, @Cambridge_Uni
Lina Khan @linakhanFTC
76K Followers 800 Following Chair, Federal Trade Commission. Retweets, Follows, & Likes ≠ Endorsements.
George Hotz 🌑 @realGeorgeHotz
248K Followers 174 Following President @comma_ai. Founder @__tinygrad__
Ate-a-Pi @8teAPi
39K Followers 2K Following self aware neuron; historian from 2130; epistemic polluter; 95 yr old man;
Atreides Management, .. @Atreidesmgmt
5K Followers 106 Following Investment firm. CIO @gavinsbaker. Crossover strategy. We want to help exceptional teams across consumer and technology. No investment advice.
Gavin Baker @GavinSBaker
126K Followers 4K Following Managing Partner & CIO, @atreidesmgmt. Husband, @l3eckyy. Barbarian, Diablo 4. Warlock, Destiny 2. No investment advice, views my own.
Institute for Progres.. @IFP
12K Followers 29 Following A think tank for accelerating scientific, technological, and industrial progress. Follow our team: https://t.co/CC0MxWfh3X
Thomas Larsen @thlarsen
46 Followers 160 Following
Nick Gabrieli @NickGabs01
4 Followers 1 Following
Caleb Watney @calebwatney
17K Followers 947 Following co-founder @IFP // the US is the R&D lab for the world and we should act like it
Rush Doshi @RushDoshi
23K Followers 4K Following Director & Sr. Fellow, CFR China Strategy Initiative • Incoming Asst Professor, Georgetown SFS • Former Deputy SD for China, Biden NSC • Author of The Long Game
Divya Siddarth @divyasiddarth
5K Followers 996 Following collective intelligence accelerationist @collect_intel, building better institutions @UniofOxford @UK AI Safety Institute, former political economist @Microsoft
Charles Martinet @Charles_Mrt
145 Followers 813 Following Intl Relations @SciencesPo. AI governance, EU foreign policy, geopolitics of tech. Currently at @Economie_Gouv. Views are my own.
@repligate banger!
until it’s clear what’s in the training mix for monster models, it’s hard to clearly say that they have zero shot capabilities this work is good - more needed
If Model A beats B in benchmarks, is it really better? Not if it trained on those benchmarks—that's an unfair edge! How can you tell if a model used benchmark data for training? 🤔 Welcome to check out our latest work: huggingface.co/papers/2404.18… (1/n)
@AndrewCritchPhD FWIW this tweet feels very…arbitrary… to me. Presumably you have a bunch of underlying beliefs informing these claims, but just flagging this as my immediate reaction. (Stylistic factors probably also contribute.)
Some thoughts from the Center for AI Policy (CAIP) on recent US Senate policy proposals on AI. Memo highlights: Two new documents from the US Senate aim to steer the future of AI safety - the "Future of AI Innovation Act" and a framework introduced by Senators Romney, Reed,…
I pre-unsupported Columbia and all Ivies by not being smart enough nor rich enough to get in
Me to the group presenting in class on a multi-billion dollar investment decision: "So, uh, do you have quantitative estimate for what the effect of [main decision] will be on [main issue]?" Them: "Look, nobody has a crystal ball; you can't provide a numerical forecast for this."
@halbritz I want this turned into a mini series starting with the DCSA investigator pulling up their new files and seeing someone listed Kim Jong Un as a foreign connection. "kim.. my cousin, youre telling me hes a dictator? no i think i would have heard"
New: A woman was denied a TS clearance for her relation to a dictator, which details suggest may be Kim Jong Un. “There is nothing about her that makes her anything less than a perfect candidate...except her family connections to a dictator," judge said cnn.com/2024/04/30/pol…
Grounds for Apple banning TikTok
A screenshot seems to show TikTok circumventing Apple's App Store fees by showing some users a link to a site for buying digital coins, instead of using IAPs (@SarahPerezTC / TechCrunch) techcrunch.com/2024/04/30/scr… 📫 Subscribe: techmeme.com/newsletter?fro… techmeme.com/240430/p41#a24…
Meta presents Better & Faster Large Language Models via Multi-token Prediction - training language models to predict multiple future tokens at once results in higher sample efficiency - up to 3x faster at inference arxiv.org/abs/2404.19737
What happened recently in AI/ML safety research (1/9) 🧵 :
Trump comes out much more forcefully in supporting Israel’s retaliation against Hamas
Trump to Hannity: "We have to let Israel complete their war on terror"
gpt2-chatbot → Generative Pretrained Transformer 2 Chatbot This is clearly a scaled up version of the Transformer 2 archtecture!! ↓
Google presents Transformer 2 - Unifies attention, recurrence, retrieval, FFN into a single module - Performs on par with Transformer w/ 20x better compute efficiency - Efficiently processes 100M context length proj: tinyurl.com/59upc7v6 abs: tinyurl.com/3nw25nz2
I'm delighted to present my NEW @CNASdc report: Beyond China's Black Box: Five Trends Shaping Beijing’s Foreign and Security Policy Decision-Making Under Xi Jinping Read it here: cnas.org/publications/r…
New episode with @JeffLadish! We chat about how hard it is to undo safety training on big language models, how to secure AI weights, and what his org Palisade Research does. YT link in reply
Episode 30 - AI Security with Jeffrey Ladish axrp.net/episode/2024/0…
I'm glad people are talking about regulation for open source AI, because depending on how compute governance goes (which I think is more important), future open source models may need to be regulated, and it will take time for legislators to get educated on the subject between…
my best guess: gpt4 knowledge + q* search reasoning = gpt2 general knowledge seems near identical to gpt4, with much better reasoning and planning capabilities. more expensive inference from tree of thought search would explain relatively slow inference and low rate limit.
irc done right is actually so much better than every other chat app experience ive had it's unreal
Trends for United States
You might like
