David Mataciunas @DeividasMat
Co-founder @ AQ22 🦾 Europe Region Lead @ Cohere for AI 💎 Chairman of the Board @ AI Association of Lithuania 🚀✨ linkedin.com/in/deividasmat… Zurich, Switzerland Joined November 2013-
Tweets3K
-
Followers215
-
Following937
-
Likes3K
Most likely explanation for gpt2-chatbot: OpenAI has been working on a more efficient method for fine-tuning language models, and they managed to get GPT-2, a 1.5B parameter model, to perform pretty damn close to GPT-4, which is an order of magnitude larger and more costly to…
and it's just a model card, the weights are all on different pages 🫡🤡
From the paper: "We find that Mistral 7B is the best performing model [among mid-size models], winning on all benchmarks and outperforming models trained specifically for the biomedical domain." 🤩
From the paper: "We find that Mistral 7B is the best performing model [among mid-size models], winning on all benchmarks and outperforming models trained specifically for the biomedical domain." 🤩
Thanks @rasbt for the signed copy. Time to brush up on some fundamentals. Making up for lost time doing half my PhD not really as an ML person. Huge respect for your education efforts to the community.
Excited to announce Med-Gemini, demonstrating a new SOTA on MedQA, multimodal and long-context abilities - arxiv.org/abs/2404.18416 I particularly want to highlight our full relabeling of MedQA, revealing that 7.4% of questions are unfit for evaluation. A short thread:
This new writeup by @cHHillee uncovered some very unexpected reasons for why we can never reach the theoretical TFLOPS advertised by accelerator vendors thonking.ai/p/strangely-ma… spoiler: it's all about the power. Make sure to read it!
Been playing with ✨B-LoRA✨, and IMO it deserves more attention key insights- ① 2 unet blocks are crucial for encoding content & style ② LoRA can be used for *implicit* style-content separation, by optimizing these blocks ③ ↑ can be done w/ 1 img ▶️ huggingface.co/papers/2403.14…
Had to give a talk to some CEOs. They knew way more about LLMs than me. Asked one of them how, he said "I check Chatbot Arena every morning" 😆 New OSGAI talk from Hao Zhang (@haozhangml ) on Chatbot Arena, seemingly the only eval anyone trusts. youtube.com/watch?v=7njmta…
Less than 12 hours ago, a mysterious new model "gpt2-chatbot" is released People are already coming up with wild use cases at GPT-4 level. 8 examples (and how to use it for free):
gpt2-chatbot is good. really good. but if this is gpt-4.5, I’m disappointed.
LLMs-as-Juries? A better way to automatically evaluate LLMs? 👨⚖️ LLM-as-a-judge refers to LLMs to evaluate the performance or quality of other LLMs. 🤔 @cohere released a new paper exploring the results of replacing a single LLM “as Judge” with multiple LLMs “Juries” where they…
Google announces Med-Gemini, a family of Gemini models fine-tuned for medical tasks! 🔬 Achieves SOTA on 10 of the 14 benchmarks, spanning text, multimodal & long-context applications. Surpasses GPT-4 on all benchmarks! This paper is super exciting, let's dive in ↓
Microsoft launches Github DEVIN! Sorry, Github Workspace!
Researchers at @ICepfl & @YaleMed teamed up to build Meditron, an LLM suite for low-resource medical settings. With Llama 3, their new model outperforms most open models in its parameter class on benchmarks like MedQA & MedMCQA. More details ➡️ go.fb.me/6vfi21
Great talk! I am constantly amazed at how similar @DrJimFan and I are in our thinking, research vision, etc. I am very interested to see what your team produces next Jim!!
Great talk! I am constantly amazed at how similar @DrJimFan and I are in our thinking, research vision, etc. I am very interested to see what your team produces next Jim!!
Demis Hassabis describes how AlphaZero, starting from scratch, became "the greatest Chess playing entity that's ever existed" in only 9 hours
CRISPR-GPT: An LLM Agent for Automated Design of Gene-Editing Experiments abs: arxiv.org/abs/2404.18021 Introduces CRISPR-GPT, a tailor-made LLM agent for automated designing of gene-editing experiments. Focuses on breaking experiment design down to a variety of substeps that…
New paper surveying multimodal LLM hallucinations: arxiv.org/abs/2404.18930 It creates a taxonomy of the varied ways hallucinations appear, with an intent to reveal causes and explain mitigation strategies. It's an educational read and an admirable effort. Hallucinations,…
Tomorrow I am giving a talk about what you should know when embarking on a research path. What is your number 1 tip?
Personal Update: I’m thrilled to share that I've joined as a Research Engineer on the phenomenal team led by @sarahookr at @CohereForAI !!!💙😍 Still can’t believe I finally get to pursue research “full-time” 🪄
Verdell Dredge @VDredge13842
83 Followers 5K FollowingMauricio Amaro L. �.. @CioAmaro
15K Followers 14K Following #IT_Thinker #strategist #speaker & #WineLover #CIO100 #HITEC50 #Cybersecurity pres. by The C-Class. #EXATEC & ex @udla_cl #IoT & #AI fanPhoebeBoswell @o3Fo5m92r3CIcIz
0 Followers 171 FollowingAlona Caballero @alon_caballe
83 Followers 5K FollowingDitagough @ditagough34982
17 Followers 216 Following Nice to meet you. My hobbies are reading, food and sports. I like cats😘 I like to meet new friends while traveling🎉🎉🎉Dora James @Doraaandus
1K Followers 3K FollowingZabir Al Nazi Nabil @PseudoEmpirical
66 Followers 323 Following Self-taught SWE, Open Source Enthusiast & Contributor, Sci-Fi Connoisseur. Interested in AGI, LLM, XAI. CS PhD Student @UCRiversideMarco D'Alia @madarco
534 Followers 341 Following Software Architect, AI Expert, Building https://t.co/MSvdYfBf9W, previously SW Manager @ Upwork, 2 times Founder, Hardware MakerEva @MiSsMaRoUa
14 Followers 583 Following Love - it's not one heart hitting another heart, but a spark of two hearts colliding together.Irmgard Lidtke @IrmgardLid
35 Followers 5K Followingomkaar @omkizzy
2K Followers 658 Following building agent infra | eng @uwaterloo | ✍️ dist sys @ https://t.co/dIQp42Qtyv for 5000+ SWEs | past @cartainc @autodeskpw018riortr0n @0t67pgqzqj
5 Followers 176 Following 【coinsrw . com 】User**me:Rom88 , P*****rd:R 66888 Bal:4,289,287,11 U.S.D.TExie Deno @DenoExie99141
57 Followers 5K FollowingCaprice Pettyjohn @CaprPettyjo
43 Followers 5K FollowingThalia Partlow @PartloThali
38 Followers 5K FollowingAndrew Trask @iamtrask
74K Followers 190 Following @openminedorg, @GoogleDeepMind ethics team, @OxfordUni phd candidate, @UN pet lab, @GovAI_, creator of #GrokkingDeepLearning, NALU, and sense2vecRena Callen @rena_call
88 Followers 5K FollowingShayne Longpre @ShayneRedford
4K Followers 997 Following PhD @MIT. Prev: @Google Brain, @apple ML, @stanfordnlp. 🇨🇦 Interests: AI/ML/NLP, Data-centric AI, transparency & societal impactPamella Entwisle @p_entwis
66 Followers 5K FollowingPorsha Bagnaschi @PorsBagnasc
87 Followers 5K FollowingAIProductDB @AIProductDB
654 Followers 2K Following AI Product Database, a site dedicated to discovering and sharing the latest and greatest AI-powered products for every use case and industry.Niesha Hellmann @HellmanNies
69 Followers 5K FollowingMadeline Arrellin @ArrelliMadelin
11 Followers 3K FollowingTodd Kueny — e/acc @techgazetteco
4K Followers 6K Following Empowering worlds where AI enriches lives, solves complex problems, and inspires continuous learning.mlecchaslayer156 @mlecchasla37448
100 Followers 3K FollowingAnnie @dionne_lamonica
3 Followers 122 Following I am a Japanese. I am both eager for love and afraid of love. I have my own career. I own a cafe, restaurant, clothing store, and electronic tradingCharlena Accosta @CharleAcco
38 Followers 5K FollowingLekisha Factor @LekishaF84482
79 Followers 5K FollowingGracie Dedier @GraciDedi
43 Followers 5K FollowingMyrtie Spaman @MyrtieS5611
81 Followers 5K FollowingMr. Jack Tung @MrJackTung
207 Followers 3K FollowingMcShywhe @shywhe934
118 Followers 4K FollowingAthena Klemen @AthenKlem
28 Followers 5K FollowingJana Filsinger @FilsingerJ20698
36 Followers 5K FollowingTempie Adamek @TempieA32157
74 Followers 5K FollowingSumayyah Whidden @SumayWhidde
32 Followers 5K FollowingPyper Totter @pyp_tott
46 Followers 5K FollowingEdward Beeching @edwardbeeching
1K Followers 70 Following Research Scientist @HuggingFace. PhD in Deep RL approaches for Robotic Navigation @INRIA.Clément ROMAC @ClementRomac
463 Followers 252 Following Research Scientist at 🤗 @huggingface, PhD. student at @FlowersINRIA. Studying how autonomous Deep RL agents 🤖 can leverage LLMs 📖 Also playing bass 🎸Quentin Gallouédec @QGallouedec
325 Followers 417 Following Research engineer @huggingface 🤗 PhD in RL Member of Stable-Baselines team: https://t.co/eX7JDWqc9FDan Jurafsky @jurafsky
27K Followers 297 Following Professor of linguistics and professor of computer science at Stanford and author of the James Beard award finalist "The Language of Food"Linus ●ᴗ● Ekens.. @LinusEkenstam
192K Followers 3K Following AI Gardener & Designer. Follow to get the latest AI trends, learn how to use AI tools to augment yourself - @bedtimestoryai @typeform @flodeskinc @thingtesting_Maxime Labonne @maximelabonne
12K Followers 437 Following Author of Hands-On Graph Neural Networks https://t.co/Q8victWUmR • Machine Learning ScientistStanford Engineering @StanfordEng
69K Followers 208 Following Our research and teaching educates leaders and helps solve global challenges.Reka @RekaAILabs
11K Followers 13 Following An AI research and product company 🫠. We are a team of scientists and engineers building state-of-the-art multimodal language models 😻David Ding @DavidDingAI
2K Followers 122 Following CEO and co-founder of @udiomusic. ex Google DeepMindudio @udiomusic
28K Followers 0 FollowingV7 @V7Labs
3K Followers 112 Following We allow you to turn your data into trustworthy AI models, or GenAI-fuelled automation workflows. Discover V7 Darwin and V7 Go, now.Leandro von Werra @lvwerra
6K Followers 310 Following Machine learning @huggingface: co-lead of @bigcodeproject and maintainer of TRL.Clément @clmt
15K Followers 2K Following AI @ Google DeepMind. Ex NVIDIA (built AI for self-driving cars + GPU data science), Twitter (started AI team), MadBits (founded+sold @ Twitter) 🇺🇸🇫🇷Eric Steinberger @EricSteinb
7K Followers 478 Following Writing code that writes code on a mission to build safe superintelligence | CEO/cofounder @magicailabsBlack Unicorn PR @BlackUnicornPR
662 Followers 762 Following 🚀🌍 Boosting #PR for #startups 📰 💻 Building #reputation in front of the publics that matter 🎯✍️ We work to give value to clients and journalists1LittleCoder💻 @1littlecoder
12K Followers 1K Following AI, ML, Open Source at - https://t.co/EKsvaArRIkTrenton Bricken @TrentonBricken
7K Followers 2K Following Trying to figure out what makes minds and machines go "Beep Bop!" @AnthropicAIManu Romero @mrm8488
21K Followers 2K Following CSO/Co-founder @maisaAI_. Head Contrib/ Ambassador🤗 @huggingface. Research 🌸@bigsciencew/@BigCodeProject | @SomosNLP_ co-founderInterconnects @interconnectsai
2K Followers 1 Following What you need to know about AI research trends, from @natolambert Wednesday mornings weekly, sometimes extra posts.Stratechery @stratechery
149K Followers 3 Following Articles and Updates from https://t.co/A7bGqyJ7db. For the author, follow @benthompson.batuhan (e/single) @isidentical
4K Followers 317 Following head of eng/silicon at @fal (fal ai labs). also a python core developer / @thePSF fellow. building the most efficient inference engine for diffusion models.Maisa @maisaAI_
3K Followers 3 Following Maisa abstracts the complexities of AI development. Powered by KPU, the most advanced reasoning system for LLMs that overcomes their intrinsic limitations.Corey Lynch @coreylynch
10K Followers 1K Following AI at @figure_robot, previously research scientist at @GoogleDeepMind.VantAI @vant_ai
993 Followers 3 Following Unlocking a new chapter in medicine by making protein interactions programmableCognition @cognition_labs
123K Followers 19 Following Makers of Devin, the first AI software engineer. We are an applied AI lab focused on reasoning, and code is just the beginning. Join us: https://t.co/tpfZwEwGiqY Combinator Universe @ycuniversecom
1K Followers 355 Following Startups. Technology. Culture. Since 2010.Lightspeed @lightspeedvp
47K Followers 2K Following Possibility grows the deeper you go. Serving bold builders of the future.Shayne Longpre @ShayneRedford
4K Followers 997 Following PhD @MIT. Prev: @Google Brain, @apple ML, @stanfordnlp. 🇨🇦 Interests: AI/ML/NLP, Data-centric AI, transparency & societal impactAlicia Curth @AliciaCurth
3K Followers 496 Following PhD student Machine Learning in Cambridge, Statistician at ❤️ In search of statistical intuition for modern ML & simple explanations for complex things 👀Sohee Yang @soheeyang_
1K Followers 428 Following PhD student/research scientist intern at @ucl_nlp/@GoogleDeepMind (50/50 split). Previously MS at @kaist_ai and research engineer at Naver Clova. #NLProc & MLTogether AI @togethercompute
27K Followers 304 Following The future of AI is open-source. Let's build together.Viktorija Mickute @VikVicariously
1K Followers 1K Following Emmy-nominated Senior Producer @AJContrast @AJEnglish/ #ONAWLA 2022 participant / Former Lithuanian TV host / @FulbrightPrg Scholar/ @mujschool gradLuc Georges 🦀 @LucSGeorges
713 Followers 319 Following Software & ML Engineer @huggingface. Curious learner.Carlos Santana @DotCSV
174K Followers 1K Following 🤖 Divulgador de Inteligencia Artificial (DotCSV) ✉️ Contacto comercial: [email protected] 📚 Enseño sobre IA en Youtube, Tiktok e InstagramNiklas Muennighoff @Muennighoff
5K Followers 323 Following @ContextualAI | Interests: AI/LLM Research & Health ❤️ | Past: @huggingface @PKU1898killian @hellokillian
23K Followers 438 Following building a universal interface between language models and computers ● https://t.co/yJVGuC0xlDMost likely explanation for gpt2-chatbot: OpenAI has been working on a more efficient method for fine-tuning language models, and they managed to get GPT-2, a 1.5B parameter model, to perform pretty damn close to GPT-4, which is an order of magnitude larger and more costly to…
and it's just a model card, the weights are all on different pages 🫡🤡
Apple OpenELM is now #1 trending model on HF! huggingface.co/models
From the paper: "We find that Mistral 7B is the best performing model [among mid-size models], winning on all benchmarks and outperforming models trained specifically for the biomedical domain." 🤩
For medicine, how do good, mid-sized, general LLMs (which may be partially trained on medical text) compare in performance to models built on medical resources like PubMed? We find that the general-purpose models now do better (Bolton, Xiong, et al. 2024) arxiv.org/abs/2404.15894
Thanks @rasbt for the signed copy. Time to brush up on some fundamentals. Making up for lost time doing half my PhD not really as an ML person. Huge respect for your education efforts to the community.
Excited to announce Med-Gemini, demonstrating a new SOTA on MedQA, multimodal and long-context abilities - arxiv.org/abs/2404.18416 I particularly want to highlight our full relabeling of MedQA, revealing that 7.4% of questions are unfit for evaluation. A short thread:
This new writeup by @cHHillee uncovered some very unexpected reasons for why we can never reach the theoretical TFLOPS advertised by accelerator vendors thonking.ai/p/strangely-ma… spoiler: it's all about the power. Make sure to read it!
Been playing with ✨B-LoRA✨, and IMO it deserves more attention key insights- ① 2 unet blocks are crucial for encoding content & style ② LoRA can be used for *implicit* style-content separation, by optimizing these blocks ③ ↑ can be done w/ 1 img ▶️ huggingface.co/papers/2403.14…
🔆From Persona to Personalization: A Survey on Role-Playing Language Agents 🔍 Dive into our comprehensive survey of RPLA technologies, their applications, and the exciting potential for human-AI coexistence. 📖 Paper: arxiv.org/pdf/2404.18231 [1/3]
Had to give a talk to some CEOs. They knew way more about LLMs than me. Asked one of them how, he said "I check Chatbot Arena every morning" 😆 New OSGAI talk from Hao Zhang (@haozhangml ) on Chatbot Arena, seemingly the only eval anyone trusts. youtube.com/watch?v=7njmta…
Less than 12 hours ago, a mysterious new model "gpt2-chatbot" is released People are already coming up with wild use cases at GPT-4 level. 8 examples (and how to use it for free):
gpt2-chatbot is good. really good. but if this is gpt-4.5, I’m disappointed.
LLMs-as-Juries? A better way to automatically evaluate LLMs? 👨⚖️ LLM-as-a-judge refers to LLMs to evaluate the performance or quality of other LLMs. 🤔 @cohere released a new paper exploring the results of replacing a single LLM “as Judge” with multiple LLMs “Juries” where they…
Google announces Med-Gemini, a family of Gemini models fine-tuned for medical tasks! 🔬 Achieves SOTA on 10 of the 14 benchmarks, spanning text, multimodal & long-context applications. Surpasses GPT-4 on all benchmarks! This paper is super exciting, let's dive in ↓
YET is the word I want to emphasize. When I hear people talk about #AI #technology, I often hear a sentence that is always structured in this manner: " AI can do xyz, but it cannot do that ..." And I have to really stop myself from saying the word yet.
Microsoft launches Github DEVIN! Sorry, Github Workspace!
Researchers at @ICepfl & @YaleMed teamed up to build Meditron, an LLM suite for low-resource medical settings. With Llama 3, their new model outperforms most open models in its parameter class on benchmarks like MedQA & MedMCQA. More details ➡️ go.fb.me/6vfi21
Great talk! I am constantly amazed at how similar @DrJimFan and I are in our thinking, research vision, etc. I am very interested to see what your team produces next Jim!!
Foundation Agent: a roadmap to build generally capable embodied AI that acts skillfully across many worlds, virtual or real. Project GR00T, the Humanoid robot foundation model, is a cornerstone for Foundation Agent. It's the North Star, the next grand challenge in our quest for…
Demis Hassabis describes how AlphaZero, starting from scratch, became "the greatest Chess playing entity that's ever existed" in only 9 hours
CRISPR-GPT: An LLM Agent for Automated Design of Gene-Editing Experiments abs: arxiv.org/abs/2404.18021 Introduces CRISPR-GPT, a tailor-made LLM agent for automated designing of gene-editing experiments. Focuses on breaking experiment design down to a variety of substeps that…
New paper surveying multimodal LLM hallucinations: arxiv.org/abs/2404.18930 It creates a taxonomy of the varied ways hallucinations appear, with an intent to reveal causes and explain mitigation strategies. It's an educational read and an admirable effort. Hallucinations,…