Nando de Freitas 🏳️🌈 @NandoDF
I research intelligence to understand what we are, and to harness it wisely. I lead a wonderfully creative AI team at @GoogleDeepMind who inspire me everyday. scholar.google.com/citations?user… London, England Joined April 2009-
Tweets10K
-
Followers97K
-
Following656
-
Likes17K
The name has changed but it remains the world's first 3D human foundation model. ChatPose (formerly PoseGPT) has a new name at the request of the #CVPR2024 reviewers. Same great work from @meshcapade and @PerceivingSys. Final #CVPR2024 version on arXiv: arxiv.org/abs/2311.18836
We're excited to announce that the Genie Team from @GoogleDeepMind will be our next invited speakers! Title: Genie: Generative Interactive Environments Speakers: @ashrewards, @jparkerholder, @YugeTen Sign up: eventbrite.co.uk/e/ucl-dark-spe… 📌 90 High Holborn 📅 Tue 30 Apr, 17:00
SnapKV: LLM Knows What You are Looking for Before Generation - Automatically compresses KV caches - Consistent decoding speed with a 3.6x increase in generation speed and an 8.2x enhancement in memory efficiency repo: github.com/FasterDecoding… abs: arxiv.org/abs/2404.14469
Can someone create a leaderboard with metrics that also measure the features Oriol highlights here: 1. multimodal performance: understanding and generating video, audio, touch, actions, proprioception. 2. long-context: long understanding and generation. I agree Gemini 1.5 Pro…
Can someone create a leaderboard with metrics that also measure the features Oriol highlights here: 1. multimodal performance: understanding and generating video, audio, touch, actions, proprioception. 2. long-context: long understanding and generation. I agree Gemini 1.5 Pro…
Microsoft presents Multi-Head Mixture-of-Experts Achieves notable improvements over the baseline MoE by using multiple MoE heads repo: github.com/yushuiwx/MH-MoE abs: arxiv.org/abs/2404.15045
Phi-3 just released by Microsoft. Three small size models (3.8B, 7B and 14B) trained on highly filtered and synthetic data. They report impressive performance since the 3.8B model (trained on 3T tokens) has MMLU of 69% matching Llama3 8B, and the 7B Phi-3 model has 75% MMLU,…
Wonderful to see this! Thanks @edwardbeeching and team 🙏 @scott_e_reed @konradzolna @SashaVNovikov @maidotgimenez @gbarthmaron
Wonderful to see this! Thanks @edwardbeeching and team 🙏 @scott_e_reed @konradzolna @SashaVNovikov @maidotgimenez @gbarthmaron
Fully agree that multimodal LLMs are the solution to robotics. This is why my team pioneered Gato (General AgenT One): arxiv.org/pdf/2205.06175… which we built over two years since 2020 to 2022. One of the most important parts of Gato was the data and its engineering pipeline.…
Fully agree that multimodal LLMs are the solution to robotics. This is why my team pioneered Gato (General AgenT One): arxiv.org/pdf/2205.06175… which we built over two years since 2020 to 2022. One of the most important parts of Gato was the data and its engineering pipeline.…
Last July I used @runwayml to do some animations to a classic 80s ballad (you know it!) over some MidJourney images of anthropomorphic flowers. (see my pinned tweet) Well, enter @HaiperGenAI - I thought I'd give them a fair shake with MJ v6 images of the same ilk. It is...cool
How to attain Multimodal World Models is a great open question in AI. The solutions will likely lead to more grounded models that interact better with people and make better physics predictions. Hopefully, they will enable scientific generalisation, but this too I feel is an open…
How to attain Multimodal World Models is a great open question in AI. The solutions will likely lead to more grounded models that interact better with people and make better physics predictions. Hopefully, they will enable scientific generalisation, but this too I feel is an open…
Llama3 was trained on 15 trillion tokens of public data. But where can you find such datasets and recipes?? Here comes the first release of 🍷Fineweb. A high quality large scale filtered web dataset out-performing all current datasets of its scale. We trained 200+ ablation…
Llama3 was trained on 15 trillion tokens of public data. But where can you find such datasets and recipes?? Here comes the first release of 🍷Fineweb. A high quality large scale filtered web dataset out-performing all current datasets of its scale. We trained 200+ ablation…
Congrats @GoogleDeepMind on shipping Gemini 1.5 Pro to public review! Upon capacity & latency testing, we have now brought Gemini 1.5 Pro up to the Arena🤖 Big improvement from Pro 1.0 to 1.5 across the board, and exceptionally strong long context understanding. Come test and…
Congrats @GoogleDeepMind on shipping Gemini 1.5 Pro to public review! Upon capacity & latency testing, we have now brought Gemini 1.5 Pro up to the Arena🤖 Big improvement from Pro 1.0 to 1.5 across the board, and exceptionally strong long context understanding. Come test and…
In addition to Llama 3, today we’re also publishing a new paper: Imagine Flash: Accelerating Emu Diffusion Models with Backward Distillation ➡️ go.fb.me/g4r584 This work from GenAI researchers is enabling new image generation features in Meta AI on @WhatsApp & web.
Toward Self-Improvement of LLMs via Imagination, Searching, and Criticizing Despite the impressive capabilities of Large Language Models (LLMs) on various tasks, they still struggle with scenarios that involves complex reasoning and planning. Recent work proposed advanced
One of the greatest minds of our times has died. This is a huge blow to the fields of philosophy, morality, consciousness and intelligence. I adored his teachings even though sometimes it took me years to get them. His ideas will live on. @danieldennett
Early 1K votes are in and Llama-3 is on FIRE!🔥The New king of OSS model? Vote now and make your voice heard! Leaderboard update coming very soon.
Early 1K votes are in and Llama-3 is on FIRE!🔥The New king of OSS model? Vote now and make your voice heard! Leaderboard update coming very soon. https://t.co/L9h9QrCkjl
Emerging AI Agent Architectures Researchers from IBM and Microsoft present this concise summary of emerging AI agent architectures. It focuses the discussion on capabilities like reasoning, planning, and tool calling which are all needed to build complex AI-powered agentic…
My full conversation with Mark Zuckerberg on the breaking Meta AI announcements, fighting in the UFC, metaverse, Ray-Ban Metas, and future technologies. But what was really touching was his thoughts on legacy and fatherhood. Timestamps: 00:00 Intro 00:37 Meta AI announcements…
AI Agents will take the abilities of LLMs to a whole new level. Here's how to build a simple agent that can use software tools like searching the web or writing and running python code (LLMs love to write @matplotlib code for you). youtube.com/watch?v=5drn2D…
AI Agents will take the abilities of LLMs to a whole new level. Here's how to build a simple agent that can use software tools like searching the web or writing and running python code (LLMs love to write @matplotlib code for you). youtube.com/watch?v=5drn2D… https://t.co/umVCQsx13K
Great analysis, approach 3 is finally in agreement! The loss scale was too low in our paper, resulting in premature termination of L-BFGS, and leading to bad fits. After fixing this we can reproduce your findings! We're also open sourcing the data in the paper, stay tuned :)
Great analysis, approach 3 is finally in agreement! The loss scale was too low in our paper, resulting in premature termination of L-BFGS, and leading to bad fits. After fixing this we can reproduce your findings! We're also open sourcing the data in the paper, stay tuned :)