AI is no longer science fiction. But who are the minds behind this revolutionary technology? Here are 10 remarkable visionaries pioneering the future of AI that you must know about:

AI Alignment: A Comprehensive Survey arxiv.org/abs/2310.19852 We break AI alignment into four categories: 1. Learning from feedback (e.g. RLHF) 2. Learning under distribution shift 3. Assurance (e.g. interpretability) 4. Governance Reply with any references we missed!

My recent talk at UCSD hosted by Jingbo Shang, covers an updated version of the RL with data material, including some new results on offline RL with LLMs for interactive dialogue agents (coming soon)! youtu.be/Iu_Uux0R0BI

Conformal Prediction + Reinforcement Learning = a winning 🏆 combination. A new University of California, Berkeley paper “Conformal Policy Learning for Sensorimotor Control Under Distribution Shifts” “This paper focuses on the problem of detecting and reacting to changes in

A robot may be unable to complete a task when limited by its morphology. Remarkably, people and some animals can get around this by not only using but also *designing* tools. We explore whether robots can also do this in our latest work! 🌐robotic-tool-design.github.io 🧵👇

The new AI Comic Factory is officially out! 👀 You can now use custom LoRA models 🤗 huggingface.co/spaces/jbilcke…

🤯 TIL there's a DARPA-funded project (thanks, Petros!) focused on migrating C++ code to Rust: "C++ to Rust Assisted Migration (CRAM) semi-automatically migrates well-designed, general-purpose C++ code into the Rust programming language." 📄cpp-rust-assisted-migration.gitlab.io

🚀 Meet David Hall, the mastermind melding NLP with next-gen AI at Stanford's Foundation Models hub! From Semantic Machines' success to Microsoft's acquisition, his journey is as layered as the AI he crafts. Catch the leader who's scripting the future of conversational AI! 🤖💬🧬

I2VGen-XL: High-Quality Image-to-Video Synthesis via Cascaded Diffusion Models - Resolution up to 1280x720 - 35M single-shot text-video pairs and 6B text-image pairs to optimize the model - Will be open-sourced arxiv.org/abs/2311.04179

Neural MMO 2.0: A Massively Multi-task Addition to Massively Multi-agent Learning doc: neuralmmo.github.io/_build/html/rs… abs: arxiv.org/abs/2311.03736

Personal read #1 of the NeurIPS Conference #neurips2023 papers (incl D&B). "RealTime QA" such as "what is Yejin Choi's citation count?" openreview.net/forum?id=HfKOI… Weekly uploaded q&a, humans w/ web search do near perfect, GPT-3 w/ search is ~60%. Potential for real-time disaster help.

OtterHD: A High-Resolution Multi-modality Model paper page: huggingface.co/papers/2311.04… present OtterHD-8B, an innovative multimodal model evolved from Fuyu-8B, specifically engineered to interpret high-resolution visual inputs with granular precision. Unlike conventional models

Video Instance Matting paper page: huggingface.co/papers/2311.04… Conventional video matting outputs one alpha matte for all instances appearing in a video frame so that individual instances are not distinguished. While video instance segmentation provides time-consistent instance

𝐓𝐫𝐚𝐧𝐬𝐟𝐞𝐫 𝐋𝐞𝐚𝐫𝐧𝐢𝐧𝐠 𝐯𝐬. 𝐅𝐢𝐧𝐞-𝐭𝐮𝐧𝐢𝐧𝐠 𝐯𝐬. 𝐌𝐮𝐥𝐭𝐢𝐭𝐚𝐬𝐤 𝐋𝐞𝐚𝐫𝐧𝐢𝐧𝐠 𝐯𝐬. 𝐅𝐞𝐝𝐞𝐫𝐚𝐭𝐞𝐝 𝐋𝐞𝐚𝐫𝐧𝐢𝐧𝐠. Most ML models are trained independently without any interaction with other models. But real-world ML uses many powerful learning

Just added ControlNet Canny to the near real-time Latent Consistency Model demo. It's much better than just img2img! Any updates to the UI parameters and prompts happen instantly. Video here at 2x speed Demo: huggingface.co/spaces/radames…

Retentive Network: A Successor to Transformer for Large Language Models BusinessIntelligence Dr. Khulood Almani | د.خلود المانع Dr. Debashis Dutta Dr. Monika Sonu I Founder Healthinnovationtoolbox Dr. Theophano Mitsa ☦️🇬🇷🇺🇸 Tony Moroney Mack - #TechTrends 💙 #TechForGood 💙 Françoise Morvan Xavier ANCELIN  Eveline Ruehlin Prof. Sally Eaves Ian Jones artgor.medium.com/paper-review-r…

اموشنال دمج

How to #RLHF for LLMs: #PPO or #DPO? Introducing #BPO (black-box prompt optimization) to align LLMs without model training. 1) ChatGPT + BPO > ChatGPT 2) GPT-4 + BPO > GPT-4 3) Vicuna + BPO > Vicuna + PPO/DPO 4) Vicuna + DPO + BPO > Vicuna + DPO arxiv.org/pdf/2311.04155…

New in Fireworks Image Generation: SSD-1B, image2image, ControlNet, and more! Read our blog: blog.fireworks.ai/new-in-firewor… Take your image-generation apps to the next level with the following new image generation features on our fast inference platform: