Stella Biderman @BlancheMinerva
Open source LLMs and interpretability research at @BoozAllen and @AiEleuther. My employers disown my tweets. She/her stellabiderman.com Joined May 2019-
Tweets12K
-
Followers15K
-
Following749
-
Likes11K
I’m presenting @aistats_conf today poster 133, check out how we can mix neuro-symbolic methods with state space models for malware, and the fu things we learn! @rea1mma @BlancheMinerva @oatesbag @BoozAllen @umbccsee arxiv.org/abs/2403.17978
I’m presenting @aistats_conf today poster 133, check out how we can mix neuro-symbolic methods with state space models for malware, and the fu things we learn! @rea1mma @BlancheMinerva @oatesbag @BoozAllen @umbccsee arxiv.org/abs/2403.17978
@tamaybes but current models don’t allocate parameters to rotary embs! this means the Chinchilla D=20*N is skewed already for the actual param counts of most models, even if it held across datasets! If we disregarded the pos. encoding params the coefficients would change
@tamaybes a super-fun arcane historical detail: Gopher (and by extension Chinchilla) use Transformer-XL style position encodings. This means they spend 20B params (Gopher) and 5B params (Chinchilla) on just rel. position encoding!
the best TTS available from 2020-2021 was done by a single unemployed guy who supported every single my little pony voice and literally nothing else the best TTS from 2022-2024 was also done by a single (different) unemployed guy w a custom-built 6x3090 rig in his basement (1/2)
#HowGPTWorks, book w/ @BlancheMinerva @drewfarris is already up to # 3 on the best sellers! We are removing all the mystery behind WTF a language model is, how they work, but in an accessible way for people without any AI/ML training. @ManningBooks manning.com/books/how-gpt-…
This seems clearly correct to me and is something I've personally experienced. Probably the easiest way to see this is true is to realize that people don't know the logical closure of their beliefs, but given time and a pencil can work many things in said logical closure out.
This seems clearly correct to me and is something I've personally experienced. Probably the easiest way to see this is true is to realize that people don't know the logical closure of their beliefs, but given time and a pencil can work many things in said logical closure out.
100%! People are really bad at understanding the logical close of their beliefs. (Proof: if they weren't, we would know if ZFC was consistent!)
100%! People are really bad at understanding the logical close of their beliefs. (Proof: if they weren't, we would know if ZFC was consistent!)
Really amazing work by the @huggingface team! Infrastructure work, including dataset work, evaluations work, and building libraries, is the single highest-leverage thing you can do in AI. This will provide dividends for the broader AI community for years to come.
Really amazing work by the @huggingface team! Infrastructure work, including dataset work, evaluations work, and building libraries, is the single highest-leverage thing you can do in AI. This will provide dividends for the broader AI community for years to come.
An essential blocker to training LLMs on public domain books is not knowing which books are in the public domain. We're working on it, but it's slow and costly... if you're interested in providing support reach out!
An essential blocker to training LLMs on public domain books is not knowing which books are in the public domain. We're working on it, but it's slow and costly... if you're interested in providing support reach out!
SSMs + long sequence analysis + malware detection with LLMs is all the buzzwords you need to decide to check our paper out, right? arxiv.org/abs/2403.17978
SSMs + long sequence analysis + malware detection with LLMs is all the buzzwords you need to decide to check our paper out, right? arxiv.org/abs/2403.17978
Training data transparency is an unambiguous win for society, but all the incentives are against companies doing it right now. We need to fix this as soon as possible.
Training data transparency is an unambiguous win for society, but all the incentives are against companies doing it right now. We need to fix this as soon as possible.
We are excited to see torchtune, a newly announced PyTorch-native finetuning library, integrate with our LM Evaluation Harness library for standardized, reproducible evaluations! Read more here: Blog: pytorch.org/blog/torchtune… Thread:
We are excited to see torchtune, a newly announced PyTorch-native finetuning library, integrate with our LM Evaluation Harness library for standardized, reproducible evaluations! Read more here: Blog: pytorch.org/blog/torchtune… Thread:
Zyphra is pleased to announce Zamba-7B: - 7B Mamba/Attention hybrid - Competitive with Mistral-7B and Gemma-7B on only 1T fully open training tokens - Outperforms Llama-2 7B and OLMo-7B - All checkpoints across training to be released (Apache 2.0) - Achieved by 7 people, on 128…
Calling all academic AI researchers! 🚨 We are conducting a survey on compute resources. We want to help the community better understand our capabilities+needs. We hope that this will help us all advocate for the resources we need! Please contribute at: forms.gle/3hEie4hj999fiS…
🚀 Introducing Pile-T5! 🔗 We (EleutherAI) are thrilled to open-source our latest T5 model trained on 2T tokens from the Pile using the Llama tokenizer. ✨ Featuring intermediate checkpoints and a significant boost in benchmark performance. Work done by @lintangsutawika, me…
I've been brain-dumping what I know about how LLMs work for several months now into an accessible general audience book! Check out the pre-release at the link.
I've been brain-dumping what I know about how LLMs work for several months now into an accessible general audience book! Check out the pre-release at the link.
(((ل()(ل() 'yoav))).. @yoavgo
46K Followers 2K FollowingDelip Rao e/σ @deliprao
46K Followers 5K Following Busy inventing the shipwreck. @Penn. Past: @johnshopkins, @UCSC, @Amazon, @Twitter ||Art: #NLProc, Vision, Speech, #DeepLearning || Life: 道元, improv, running 🌈near @nearcyan
46K Followers 882 Following https://t.co/IdaJwZJCXm partner @ https://t.co/9g1MIgjiqc dms openSynthLabs @synth_labs
12K Followers 43 Following AI Aligned with Your Vision. We’re doing cutting edge research for transparent, auditable AI alignment.MMitchell @mmitchell_ai
80K Followers 1K Following Interdisciplinary researcher focused on shaping AI towards long-term positive goals. ML & Ethics. Same content in the Sky, Threads, & the Prehistoric ElephantEric Jang @ericjang11
69K Followers 3K Following physical AGI at 1X. Author of "AI is Good for You" https://t.co/eFg4WXhg0pMiles Brundage @Miles_Brundage
43K Followers 10K Following Policy research at @openai. I mostly tweet about AI, animals, and sci-fi. He/him. Views my own.Sasha Luccioni, PhD �.. @SashaMTL
19K Followers 4K Following AI & Climate Lead @HuggingFace, Board Member of @WiMLworkshop, Founding Member of @ClimateChangeAI. @TEDTalks speaker. She/her/Dr/ 🦋Horace He @cHHillee
24K Followers 449 Following Working at the intersection of ML and Systems @ PyTorch "My learning style is Horace twitter threads" - @typedfemaleclem 🤗 @ClementDelangue
91K Followers 5K Following Co-founder & CEO @HuggingFace 🤗, the open and collaborative platform for AI buildersTalia Ringer 🟣 �.. @TaliaRinger
26K Followers 6K Following Professor, @plfmse, @IllinoisCS! Proof Automation. @SigplanM & CCF Founder. Israeli-American for peace, equality, & justice. They/היא, ND, bi. די לכיבושRivers Have Wings @RiversHaveWings
31K Followers 226 Following AI/generative artist. Writes her own code. Absolute power is a door into dreaming.merve @mervenoyann
56K Followers 4K Following open-sourceress at @huggingface 🧙🏻♀️ proud mediterrenean 🍋 I do TL;DR on ML papersJulien Chaumond @julien_c
47K Followers 1K Following Co-founder and CTO at @huggingface 🤗. ML/AI for everyone, building products to propel communities fwd. @Stanford + @PolytechniqueThomas Wolf @Thom_Wolf
68K Followers 4K Following Co-founder and CSO @HuggingFace - open-source and open-scienceSoumith Chintala @soumithchintala
187K Followers 887 Following Cofounded and lead @PyTorch at Meta. Also dabble in robotics at NYU. AI is delicious when it is accessible and open-source.Rosanne Liu @savvyRL
33K Followers 969 Following Cofounded & running @ml_collective. Host of Deep Learning Classics & Trends. Research at Google DeepMind. DEI/DIA Chair of ICLR & NeurIPS. Writing https://t.co/IbycyGfnDRJack Clark @jackclarkSF
68K Followers 5K Following @AnthropicAI, ONEAI OECD, co-chair @indexingai, writer @ https://t.co/3vmtHYkaTu Past: @openai, @business @theregister. Neural nets, distributed systems, weird futuresPercy Liang @percyliang
50K Followers 408 Following Associate Professor in computer science @Stanford @StanfordHAI @StanfordCRFM @StanfordAILab @stanfordnlp | cofounder @togethercompute | PianistTanishq Mathew Abraha.. @iScienceLuvr
55K Followers 1K Following PhD at 19 | Founder and CEO at @MedARC_AI | Research Director at @StabilityAI | @kaggle Notebooks GM | Biomed. engineer @ 14 | TEDx talk➡https://t.co/xPxwKTq6QbHari Nair @harisnair
302 Followers 1K Following Enabling Digital Transformation within Enterprises l Toastmaster. Views are personal.Heegon Jin @jineussw
24 Followers 577 Following Exploring the frontiers of AI 🤖 Research Scientist @ NCSOFT working on Machine Translation and NLP | #NLProc #AIAshwin Jayaprakash @ashwinjay
206 Followers 4K Following Falling into the future at light speed. (Any opinions expressed are my own)Cam Howe (e/acc) @camhowe1729
4 Followers 260 Following full-time techbro, part-time anon/undergrad. love explaining tech stuff.Ahmed Mujtaba @ahmedmujii
24 Followers 256 Followingokeblay @okeblay
8 Followers 96 FollowingNitish ⚡️ @nitishmutha
3K Followers 333 Following Co-founder and CTO @GenieAI - Building the world’s best AI Legal Assistant. @UCL alum.Sujan Kumar @kumarsujan
43 Followers 444 Following Sr. ML Scientist at AWS AI Labs @AmazonScience @awscloudAustin Kwoun @kwoun_austin
0 Followers 10 FollowingDaniel @dabronofstock
13 Followers 78 Followingsteady rockin all nig.. @sonikudzu
2K Followers 862 Following little whirl in her crackpot era. reticulating sage splines. cream of stonsciousness. mad hatter. show me my opponent “in my opinion:”dooniek @doonielk
33 Followers 369 Followingdwrodri @__dwrodri
201 Followers 1K Following Software performance, Machine Learning, Prob/Stats, finance/econ, electronic music, lean/small tech businesses. Opinions are most certainly definitely my own.Electronicsseeker @libertarian108
39 Followers 3K FollowingJim Auwerda @JimAuwerda
93 Followers 1K FollowingJulien Borderieux @J_Borderieux
5K Followers 6K FollowingMohammed _h_mno3 @Slalucynkdia
531 Followers 23 Following Palestine journalist, media director at ✈️ Rafah border crossing( and there hearts will be reassured in the remembrance of God)Jacob Somer @jacob_somer_
666 Followers 4K Following AI Enthusiast & Software Engineer 💻 Building intelligent systems that make a difference.Andrei Savu @andreisavu
7K Followers 6K Following Actions Speak Louder Than Words | Awareness Above Wishful ThinkingGmail Accounts 🇺�.. @accounts_g1158
67 Followers 688 Following #Bitcoin #USDT #Ethereum #Payoneer #Direct_Bank_Transfer #PayPalShadab Choudhury @ShcChy
105 Followers 409 Following MS CS @UMBC | Accessibility and Multimodal DL | every day I edge closer to weebpfp anonpoastingKristina Terech @kristina_terex
5 Followers 37 Following Software writer at TechRadar. She/Her. Views my own.Brian Roach @itsbrex
653 Followers 2K Following Technical Product Manager | MBA, AI+ML #OpenToWork | 🪩 @UCSD, @USC, @UCBerkeleyTiezhen WANG @Xianbao_QIAN
928 Followers 369 Following Engineer at HuggingFace, ex-Googler on TFLite / micro. Ideas are my own.Wilf Rosenbaum @WilfRosenbaum
104 Followers 962 FollowingSrmouse @mousebars
269 Followers 5K FollowingBill Qian @billqian_uae
1K Followers 4K Following CIO of Phoenix Group| Chairman of Cypher Capital | CIO of https://t.co/fR7qMZOHi1 | Board Member of TON Foundation | Ex-Global Head of M&A/Labs- BinanceCan Yaman @CanYaman_21
13 Followers 578 Following Can Yaman is a Turkish actor who was born on November 8, 1989 in Istanbul, Turkey. He is 6'2". He won a Golden Butterfly Award for Best Actor in a Romantic.Nikita @nikitavoloboev
4K Followers 7K Following Make @LearnAnything_ Learn in public: https://t.co/GbFvuErkYn macOS course: https://t.co/JdbJWru6zG https://t.co/94R8ER7K2h https://t.co/ROkqhyhpEKshubhpa𝐭ni.eth @PatniShubh
805 Followers 3K Following buidling @resmeai | 5x hackathon 🏆 and top writer on mediumHamzé @Hamzeml
510 Followers 5K Following A Humanist Technologist, AI optimist, CTO @gowelcomeplace, #inclusive_economy #AI #machinelearning #tech4good #edtechWes George @vvxgeorge
129 Followers 539 Following Technologist, Skeptic, Human. solutions @recursal_AI. I fight for the users.Nextt @_Nextt_
20 Followers 11 FollowingShashank Shekhar @sshkhr16
2K Followers 1K Following Scale Maximalist. Opinions my own, ofc Previously: AI Research @MetaAI @uofg @vectorinst @_NextAI @iiscbangaloreDean Clark @DeanCla88922559
148 Followers 1K Following Disabled part-time student, registered for artificial kidney trials.Christian Miranda @cmoryah
212 Followers 454 FollowingJen Iofinova @oohaijen
238 Followers 699 Following Once a mathematician, twice an immigrant. PhD student @ISTAustria building efficient, trustworthy ML. Formerly: software engineer @Google, teacher, wall street.A. J. Kübler 👨�.. @ajkuebler
62 Followers 2K FollowingZhiyong Wang @Zhiyong16403503
437 Followers 3K Following Visiting Ph.D. student at Cornell University. Ph.D. candidate at CUHK. Working on bandits and reinforcement learning theory.Syner @renchengyuan
230 Followers 3K FollowingJprwg @jprwg
513 Followers 3K Following No idea if that's true, but it is at least faintly entertaining.Hpremium @web3nam3
805 Followers 3K Following https://t.co/Tes5ZFnfVs • https://t.co/NuhiRgwvTP https://t.co/3H4X5XEq21 •https://t.co/oOCFfDThZ2 • https://t.co/wMpswOH3Xa • https://t.co/cW7uHNvbfy • https://t.co/VYahFk94rN •https://t.co/Gik8R81APV• https://t.co/Uvx07c8pI2 • https://t.co/yp6o2BXYZH•https://t.co/0V7mofFuIl •📈(((ل()(ل() 'yoav))).. @yoavgo
46K Followers 2K FollowingDelip Rao e/σ @deliprao
46K Followers 5K Following Busy inventing the shipwreck. @Penn. Past: @johnshopkins, @UCSC, @Amazon, @Twitter ||Art: #NLProc, Vision, Speech, #DeepLearning || Life: 道元, improv, running 🌈Lucas Beyer (bl16) @giffmana
56K Followers 447 Following Researcher (Google DeepMind/Brain in Zürich, ex-RWTH Aachen), Gamer, Hacker, Belgian. Mostly gave up trying mastodon as [email protected]SynthLabs @synth_labs
12K Followers 43 Following AI Aligned with Your Vision. We’re doing cutting edge research for transparent, auditable AI alignment.MMitchell @mmitchell_ai
80K Followers 1K Following Interdisciplinary researcher focused on shaping AI towards long-term positive goals. ML & Ethics. Same content in the Sky, Threads, & the Prehistoric ElephantEric Jang @ericjang11
69K Followers 3K Following physical AGI at 1X. Author of "AI is Good for You" https://t.co/eFg4WXhg0pMiles Brundage @Miles_Brundage
43K Followers 10K Following Policy research at @openai. I mostly tweet about AI, animals, and sci-fi. He/him. Views my own.Sasha Luccioni, PhD �.. @SashaMTL
19K Followers 4K Following AI & Climate Lead @HuggingFace, Board Member of @WiMLworkshop, Founding Member of @ClimateChangeAI. @TEDTalks speaker. She/her/Dr/ 🦋Horace He @cHHillee
24K Followers 449 Following Working at the intersection of ML and Systems @ PyTorch "My learning style is Horace twitter threads" - @typedfemaleAnthropic @AnthropicAI
265K Followers 26 Following We're an AI safety and research company that builds reliable, interpretable, and steerable AI systems. Talk to our AI assistant Claude at https://t.co/aRbQ97uk4d.clem 🤗 @ClementDelangue
91K Followers 5K Following Co-founder & CEO @HuggingFace 🤗, the open and collaborative platform for AI buildersTalia Ringer 🟣 �.. @TaliaRinger
26K Followers 6K Following Professor, @plfmse, @IllinoisCS! Proof Automation. @SigplanM & CCF Founder. Israeli-American for peace, equality, & justice. They/היא, ND, bi. די לכיבושRivers Have Wings @RiversHaveWings
31K Followers 226 Following AI/generative artist. Writes her own code. Absolute power is a door into dreaming.Julien Chaumond @julien_c
47K Followers 1K Following Co-founder and CTO at @huggingface 🤗. ML/AI for everyone, building products to propel communities fwd. @Stanford + @PolytechniqueThomas Wolf @Thom_Wolf
68K Followers 4K Following Co-founder and CSO @HuggingFace - open-source and open-scienceSoumith Chintala @soumithchintala
187K Followers 887 Following Cofounded and lead @PyTorch at Meta. Also dabble in robotics at NYU. AI is delicious when it is accessible and open-source.Hugging Face @huggingface
347K Followers 188 Following The AI community building the future. https://t.co/VkRPD0VKaZ #BlackLivesMatter #stopasianhateRosanne Liu @savvyRL
33K Followers 969 Following Cofounded & running @ml_collective. Host of Deep Learning Classics & Trends. Research at Google DeepMind. DEI/DIA Chair of ICLR & NeurIPS. Writing https://t.co/IbycyGfnDRhackerfantastic.x @hackerfantastic
103K Followers 4K Following Co-Founder @myhackerhouse cyber security assurance & hacker training ~ ISBN9781119561453 ~ a book on professional hacking. Offensive Lua project.Ian Coldwater 📦�.. @IanColdwater
106K Followers 1K Following Kubernetes SIG Security co-chair, container escape artist, goose in the mainframe. They/them. Legacy verified. Stay punk 🏴Tom McCoy @RTomMcCoy
3K Followers 483 Following Assistant professor @YaleLinguistics. Studying computational linguistics, cognitive science, and AI. He/him.Ruochen Zhang @ruochenz_
327 Followers 1K Following PhDing @Brown_NLP & @health_nlp, working on multilingual NLP. Prev: Undergrad @sutdsg, she/herSebastian Majstorovic @storytracer
2K Followers 812 Following Digital Historian & Data Consultant | https://t.co/fev0QjCWjp | https://t.co/yqa5eIfpTu | Co-Founder @sucho_orgDavid @DavidSHolz
54K Followers 5K Following founder @midjourney, prev founder leap motion, nasa, max planckAshish Vaswani @ashVaswani
19K Followers 2K FollowingLucia Quirke @lucia_quirke
215 Followers 85 Following Neural network interpretability researcher at EleutherAIChris Ociepa @ChrisOciepa
27 Followers 68 Following Specialized in the design and development of scalable distributed systems with BigData & AI. Passionate about hacking and training LLMs. A huge fan of astronomypleias @pleiasfr
242 Followers 1 FollowingBlinkDL @BlinkDL_AI
7K Followers 92 Following RWKV = 100% RNN with GPT-level performance. https://t.co/TkdxOJSFWX and https://t.co/86DzS6arA0Kevin Klyman @kevin_klyman
3K Followers 3K Following AI policy @StanfordHAI + avoiding war with China @BelferCenter. Words in @ForeignPolicy @TechCrunch et al. Ex @UNGlobalPulse @BanKillerRobots @hrwLogan Kilpatrick @OfficialLoganK
92K Followers 2K Following Lead product for @Google AI Studio and working on the Gemini API, helping developers build with AI, my views!Gowthami Somepalli @gowthami_s
6K Followers 981 Following Grad student @UMDCS. Past: @AIatMeta, @AmazonScience, @IITMadras. Currently working on #Diffusion and #Multimodal understanding. GPU poor. She/her.Ross @rpoo
25K Followers 1K FollowingAspen @aspenkhopkins
558 Followers 195 Following PhD Student advised by @aleks_madry @MIT_CSAIL interested in data and ml. On bluesky @ dataspen.bsky.socDimitri von Rütte @dvruette
710 Followers 171 Following Studies @ETH_en, Machine Learning @DeepJudgeAILennart Heim @ohlennart
3K Followers 823 Following huh? | AI (Compute) Governance @GovAI_ | Also @EpochAIResearch |Conference on Languag.. @COLM_conf
2K Followers 6 Following https://t.co/GhGCMEoa4A Abstract submission: March 22, 2024Conglong Li @conglongli
133 Followers 61 Following Senior Researcher @Microsoft DeepSpeed team, working on deep learning systems. @SCSatCMU PhD, @RiceCompSci BS+MS. Views are my own. English/Chinese/Japanese.Dimitris Papailiopoul.. @DimitrisPapail
12K Followers 981 Following prof @ wisconsin; thinking about transformers; learning in context; babas of Inez LilyYi Tay @YiTayML
29K Followers 97 Following chief scientist / cofounder @RekaAILabs 🫠 past: research scientist @google brain 🤯 currently learning to be a dad 🍼Sophie @lebrechts
905 Followers 839 Following COO @allen_ai formerly AI/ML @Apple, SVP Strategy & Ops https://t.co/4Z5RuqSEkZ, PhD from @BrownUniversity, post-doc @CarnegieMellonIz Beltagy @i_beltagy
2K Followers 422 Following Cofounder @SpiffyAI, Research Lead building OLMo at @allenai_org, formerly @UTCompSci PhD.Mechanical Dirk @mechanicaldirk
550 Followers 244 Following Principal Engineer at @allen_ai. Engineering Lead of the OLMo project.Deepak Narayanan @deepakn94
1K Followers 1K Following Research Scientist at @nvidia. Interested in the intersection of Computer Systems and ML. Occasionally tweet about sports. Views are my own.Briana Vecchione @brianavecchione
3K Followers 1K Following Researcher @datasociety • PhD @CornellCIS • AI auditing/accountability • Prev. @Meta Fellow @MicrosoftNY @spotifyresearch @Mozilla • 🏳️🌈PicoCreator (🇸🇬.. @picocreator
2K Followers 164 Following Builds Attention-Free Transformer (https://t.co/YL7CbNYKBs) from scratch - CEO @ https://t.co/kQHiGtzJWr Also built k8s tools, uilicious & GPU.js (https://t.co/OIfnI1EPU7)michele benetti @michelebenben
65 Followers 600 FollowingBirchlabs @Birchlabs
4K Followers 172 Following ML Engineer at Anlatan (@novelaiofficial). co-author of HDiT (Hourglass Diffusion Transformers). works on diffusion models and LLMs. 日本語を勉強してる。Aspiringspike @Aspiringspike
26K Followers 526 Following I am MTGO user and former preview card getter Aspiringspike https://t.co/pn7Z5hUMcz he/him business inquiries [email protected]Asma Ghandeharioun @ghandeharioun
2K Followers 489 Following Research Scientist @GoogleAI working on ML interpretability & human-centered AI, PhD from @MITBen Rubinstein @bipr
2K Followers 847 Following ML & Privacy Prof @cis_unimelb. Deputy Dean (Research) @engunimelb. Prev @MSFTResearch, @Berkeley_EECS. He/him. 🇦🇺Sasho Nikolov (thesas.. @thesasho
4K Followers 432 Following Associate professor at U of T. Computer science and math research: (differentially) private data analysis, geometry, discrepancy, optimization.Nalini Joshi @monsoon0
16K Followers 699 Following mathematician, wife, mother, Professor, addicted to mathTom Gur @TomGur
4K Followers 281 Following Associate Professor of Theoretical Computer Science @Cambridge_Uni. My research is in Complexity Theory and Quantum Computing.Niloofar (Fatemeh) @I.. @niloofar_mire
5K Followers 1K Following Postdoc @uwcse-@uwnlp, Ph.D. from @ucsd_cse /Privacy, ML, NLP, @winlpworkshop chair, @MSFTResearch - Semantic MachinesJohann Rehberger @wunderwuzzi23
3K Followers 628 Following Hacking neural networks so that we don’t get stuck in the matrix. Red Team Director @ Electronic Arts. Entrepreneur. Builder and Breaker. Opinions are my own.Sharon Goldman @sharongoldman
6K Followers 2K Following Reporter on the AI beat, @FortuneMagazine Formerly @venturebeat [email protected] | Signal: sharongoldman.43 (for news tips only, no pitches)nathan lile @ ICLR '2.. @NathanThinks
2K Followers 891 Following ceo/cofounder @ https://t.co/bDd3J4KOJH (we're hiring!) #GenerativeAI recurrent rabbit hole victim. swims in data lakes & pools. nothing great is easy.Daniel van Strien @vanstriendaniel
3K Followers 1K Following Machine Learning Librarian @huggingface 🤗 | Championing Open Science & ML | Sharing the latest ML datasets 🌟 | Tips for mastering the HF HubJoseph Thacker @rez0__
49K Followers 885 Following the promptfather. christian. hacker. hobby jogger. principal ai engineer @appomnisecurity.Laura O'Mahony @_lauraaisling
240 Followers 476 Following PhD Candidate SFI CRT in foundations of data science in UL 📚 just following the gradient of interesting 📈Nomic AI @nomic_ai
14K Followers 50 Following Building explainable and accessible AI https://t.co/bbYqCdL8vQBack when I'd play Minecraft, I liked making modpacks more than playing the game. I think this directly led to me curating and building datasets to train models with lol
@vishalmisra why would an LLM not be able to accept input or examples? to recursively self improve it will need to accept its last N outputs as input. did you just find a contrived, isolated case where improvement is not possible because of the constraints you injected?
If the LLM fills out these rows using no external input or prompt, then using a simple entropy argument one can show that the total information content in the matrix cannot increase. (2/n)
I’m presenting @aistats_conf today poster 133, check out how we can mix neuro-symbolic methods with state space models for malware, and the fu things we learn! @rea1mma @BlancheMinerva @oatesbag @BoozAllen @umbccsee arxiv.org/abs/2403.17978
Inspired by @_albertgu recent works in state space models, can we merge them with #VSA and #HRR for our long sequence classification needs in #malware? Our HGConv says yes! With some interesting results on pros/cons. Lead by @rea1mma w/ @BlancheMinerva Tim Oates & Jim Holt!
Finally, wanted to give a shout out to the “Do ImageNet Classifiers Generalize to ImageNet?” paper by @beenwrekt, @BeccaRoelofs, @lschmidt3 and @Vaishaal which was a huge inspiration for this work and a longtime favorite of mine. We learned a lot of lessons from them.
@GuilleAngeris I think this is literally correct (suitably interpreted)
@1a3orn Was meaning to make a claim about the substance here, not what everyone in the AI risk community believes — agree some people do worry about existing systems directly, I disagree with them and think OS has been positive so far
@daniel_271828 A feature/bug by design of this API-based method is the question of who has access. I've only been able to get into research via open models. Folks like me who didn't have the right credentials or connections could be disenfranchised. What if I'm publicly critical of the lab?
@BlancheMinerva Two of my recent papers have the sentence "We filter all questions from ARC that do not have four choices" 😔
@maksym_andr @ncmeade Yeah I think trying to dis-entangle which part of the procedure is causal is extremely tough. The pythia framework by @BlancheMinerva is fantastic for running such studies!
the amount the (then) undergrads and masters students ive worked with are rooting for me re: the faculty search is truly 🥹
@aidanogara_ @daniel_271828 nnsight is very cool but does not solve the “making it easy to reverse engineer” issue afaik?
@daniel_271828 @BogdanIonutCir2 Of course there's current demand for interp access to closed source models. You don't think researchers want to run their interp experiments on GPT-4? OpenAI doesn't even bother to even maintain consistent, reproducible behavior out of their current APIs. Counting on them to…
@daniel_271828 @nonagonono thing is this isn't an "API" anymore though, it's (in the limit of hostile actor wanting valuable weights) a facility ppl go to and work at, with their pockets searched for USB drives and even short of that, it's a sandbox w/ huge friction for researchers, headache for $LLMCO
@daniel_271828 @idavidrein is that a serious proposal or are you memeing
@daniel_271828 Realistically, who gives you access to weights via API though? As far as I can see, even getting a few top logits is a problem, let alone weights. Additionally, a lot of research requires you to make a ton of custom, unforeseen changes and APIs are typically super bad for that.
@daniel_271828 @SecondA16110022 So your argument hinges on an imaginary future API that companies would have no incentive to create?
@BlancheMinerva Yeah, I was chatting with @soldni and I had no idea how awesome Pythia was. Pythia is way more than model weights.
ty we try 😭
@BlancheMinerva Yeah, I was chatting with @soldni and I had no idea how awesome Pythia was. Pythia is way more than model weights.
@WenhuChen In theory i guess meta could demand you take down your work, since you don't have a license (since you're not following the license requirements)
Or just bad writers. You probably want an LLM to delve into my writing and make suggestions. I’m like 90% typos and run on sentences.