🧵[2/9] If you wanna learn about DROID the fun way, here’s a video explaining what the heck is going on. Otherwise, keep on scrolling! youtu.be/xqCJkTjAXDE

Martin Shkreli (e/acc) I could be wrong here, would love to be checked on my thinking: Inflection AI spends about $1B on a big H100 cluster. Next year, the next company "I2" spends $1B on a B100 cluster, which is 4x faster for same cost. At 25% of SOTA speeds, you can think of the value of the cluster

Nvidia announces AI nurses: 90% cheaper than humans, and “outperform human nurses at bedside manner” Why this matters: 1) America has 5 million nurses (6th most common job) 2) If AIs can outperform nurses, your job isn't safe for long 3) BUT compared to many other jobs, I

Very simple idea, but can turn out to be very useful if used correctly.

x.com/i/broadcasts/1…

“By combining Accenture's expertise in disruptive technology with Sanctuary AI’s industry-leading robotics, we can help some of the biggest companies in the world manage this change and provide the best solutions for its clients.” Our CEO Geordie Rose on today’s announcement

@TeslaBotJournal Brett Adcock I know there are others here on X that think Figure and their CEO is a scam, but I'm seeing a well run company. They have $700M+ in the bank, OpenAI partnership & initial customers. The only looming problem for them is scale manufacturing. They are going to go through

BREAKING: Sanctuary AI, one of the leading humanoid bot companies, just announced a blockbuster partnership with Magna International - one of the largest companies in the automotive supply sector.  And this partnership is wide ranging. Not only is Manga providing additional

📢 Code release for 𝗗𝗦𝗜𝗡𝗘 (#CVPR2024 - Oral) DSINE gives you surface normal prediction (+ uncertainty) in real-time. We have released the code for training, testing, and running real-time demos. Try it yourself! github.com/baegwangbin/DS…

How overfit are popular LLMs on public benchmarks? New research out of @scale_ai SEAL to answer this: - produced a new eval GSM1k - evaluated public LLMs for overfitting on GSM8k VERDICT: Mistral & Phi are overfitting benchmarks, while GPT, Claude, Gemini, and Llama are not.

Good Morning Yokohama! My group and I will be presenting several papers this week, on imitation learning and VLMs for robot manipulation. See below for a preview! Looking forward to seeing everybody and their robots 😀🤖 #ICRA2024

This is wild. OpenAI just dropped ChatGPT-4o and it will completely change the AI assistant game. 10 wild examples: 1. Visual assistant in real-time

We are THRILLED to have been ranked 2nd in the world and 1st in the UK and Europe in this year's QS QS World University Rankings! 🎊🥳 #QSWUR Thank you to everyone in #OurImperial community for making this result possible! 💙

Actually, really liked the Apple Intelligence announcement. It must be a very exciting time at Apple as they layer AI on top of the entire OS. A few of the major themes. Step 1 Multimodal I/O. Enable text/audio/image/video capability, both read and write. These are the native

🌟New paper!🌟 "R+X: Retrieval and Execution from Everyday Human Videos" By using a VLM for retrieval and in-context IL for execution, robots can now learn from unlabelled videos of humans performing tasks. No need to label and train; just *retrieve* and *execute*! More 👇

Excited to introduce Diffusion Augmented Agents (DAAGs)✨. We give an agent control of a diffusion model, so it can create its own *synthetic experience*.🪄 The result is a lifelong agent that can learn new reward detectors and policies, much more efficiently. Here's how. 👇

Today at #CoRL2024, we present: "MILES: Making Imitation Learning Easy with Self-Supervision" We show that ⭐️self-supervision⭐️ enables learning complex tasks from: - 1 demo - 1 reset - No prior object knowledge Come to our poster (session 3)! robot-learning.uk/miles More 👇

Many of you asked for code & weights for π₀, we are happy to announce that we are releasing π₀ and pre-trained checkpoints in our new openpi repository! We tested the model on a few public robots, and we include code for you to fine-tune it yourself.

OpenAI has recently announced they are planning to release an open source and open-weight reasoning LLM. 🤔 Whether a company decides to open source a model or keep it proprietary is highly dependent on their business model. So why could it make sense for OpenAI to do this and

Last month we announced Gemini Robotics (GR) and Gemini Robotics-ER (GR-ER). GR-ER is a powerful VLM specialised for spatial understanding, including detecting object poses in 2D/3D, pointing, and even *predicting grasp poses*. Take a look at this demo. Details below. 🧵