Mech Interp

Cluely is out. cheat on everything.

Everything new is also old. This from my 1984 PhD thesis: "AI is an experimental science, yet the complexity of its programs and problem domains often makes the interpretation of results very difficult. Programs often contain so many components and parameters that limitations on computer time and the sheer number of possibilities make it impossible to experimentally evaluate how each contributes to performance." Then I argued, just as I do today, for careful empirical studies in simplified settings that enable better scientific understanding.

I seen footage, I stay noided

When will AI systems be able to carry out long projects independently? In new research, we find a kind of “Moore’s Law for AI agents”: the length of tasks that AIs can do is doubling about every 7 months.

@e_p_armstrong @andrewversecast We’ll be talking again 2pm next Monday. Would love to have you with us again.

@giordanorogers @andrewversecast Was really fun talking with you and Andres in this space! Question: are you planning any like this in the future? Would love to be a part of them too 🙂

Come join me and @andrewversecast talking building with AI & LLMs

@PeterBowdenLive I appreciate it Peter. Excited to hear about your project’s progress next time we get a chance to connect.

@giordanorogers Have a great space tomorrow. Can’t be there due to family visit but look forward to another session.

A think-tank for anyone building AI or LLM-based projects. Come to learn, share, or just to join a conversation about building generative AI applications. 2/19: Me and @andrewversecast twitter.com/i/spaces/1RDGl…

This is so understatedly impressive. The L3GO image is realistic, aesthetic, and mathematically accurate. And with a prompt as simple as “a chair with five legs”.

L3GO Language Agents with Chain-of-3D-Thoughts for Generating Unconventional Objects paper page: huggingface.co/papers/2402.09… Diffusion-based image generation models such as DALL-E 3 and Stable Diffusion-XL demonstrate remarkable capabilities in generating images with realistic and unique compositions. Yet, these models are not robust in precisely reasoning about physical and spatial configurations of objects, especially when instructed with unconventional, thereby out-of-distribution descriptions, such as "a chair with five legs". In this paper, we propose a language agent with chain-of-3D-thoughts (L3GO), an inference-time approach that can reason about part-based 3D mesh generation of unconventional objects that current data-driven diffusion models struggle with. More concretely, we use large language models as agents to compose a desired object via trial-and-error within the 3D simulation environment. To facilitate our investigation, we develop a new benchmark, Unconventionally Feasible Objects (UFO), as well as SimpleBlenv, a wrapper environment built on top of Blender where language agents can build and compose atomic building blocks via API calls. Human and automatic GPT-4V evaluations show that our approach surpasses the standard GPT-4 and other language agents (e.g., ReAct and Reflexion) for 3D mesh generation on ShapeNet. Moreover, when tested on our UFO benchmark, our approach outperforms other state-of-the-art text-to-2D image and text-to-3D models based on human evaluation.

@andrewversecast One time I tried using the advice that threatening the LLM leads to better response. I tried it and immediately felt wrong and a little worried.

🚨 BREAKING: Nvidia just released Chat with RTX, an AI chatbot that runs locally on your PC. It can summarize or search documents across your PC's files and even YouTube videos and playlists. The chatbot runs locally, meaning results are fast, you can use it without the internet, and the user's data stays private. New day, new chatbot. Let's go.

The Copilot Super Bowl commercial really makes you feel a part of something big. Get’s me hyped up just rewatching it. Despite the hate, AI copilots & assistants are here to stay. Hopefully, people will come to view them as collaborators that make the journey more fun.

If crypto is banned because of energy consumption... Would that be valid? Or would it just be an excuse to protect the US Dollar?

Andrew Ng: “There is no definition for what is conscious or not.”

@Jalvarez0907_ Become an expert on whatever excites you now. The adoption curve is gonna keep getting more competitive.

We are currently at a point where we can all grow and learn LLM as it develops. Really interesting being part of this wave of development

"There's no one walking around with 4 years of LLM experience… So if you can learn fast, you're gonna be at the same level as everybody else."

More Agents Is All You Need It seems like just scaling up the agents in your system is a solid way to make progress. Which makes sense. An agent is essentially just a tool with memory. It's similar to a 100 person company having an advantage over a 10 person company. If you divide the work wisely and precisely, then the more agents the better. 🔗: arxiv.org/abs/2402.05120