Links for 2023-04-10
Towards ML-enabled cleaning robots https://ai.googleblog.com/2023/04/towards-ml-enabled-cleaning-robots.html
The results from this work demonstrate that complex visuo-motor tasks such as table wiping can be reliably accomplished without expensive end-to-end training and on-robot data collection. The key consists of decomposing the task and combining the strengths of RL, trained using an SDE model of spill and crumb dynamics, with the strengths of trajectory optimization.
Agentized LLMs will change the alignment landscape https://www.lesswrong.com/posts/dcoxvEhAfYcov2LA6/agentized-llms-will-change-the-alignment-landscape
“An intriguing trend in AI 🤖: “Models all the way down” (aka "stacking"). Have models invoke other models, then watch as emergent intelligence develops” https://twitter.com/mathemagic1an/status/1645096275392745477
“The demand for GPUs to train AI models is causing a major shortage right now, with AWS, Microsoft, Google, and Oracle all limiting GPU availability to customers. Some customers are reporting monthslong wait times for GPUs.” https://twitter.com/Old_Samster/status/1644410302514208768
🤖Multi-Action Agent: With a small little update to the AgentExecutor, it's now possible to have an agent output MULTIPLE agent actions to take https://python.langchain.com/en/latest/modules/agents/agents/custom_multi_action_agent.html
“What if AI agents could write their own tools/plugins? Introducing 🧰 Toolkit: the easiest way to create & discover AI Plugins for 🦜📎LangChain and ChatGPT. Just describe a plugin in natural language, and get working LangChain Code.” https://twitter.com/NicolaeRusan/status/1644120508173262853
“🔥Two AI Agents - Role-playing to create a @Gradio application by themselves 🤯 It's using CAMEL - Communicative Agents for “Mind” Exploration of Large Scale Language Model Society” https://twitter.com/1littlecoder/status/1643725760228630533
Do you want to chat with your long PDF docs? And do you know that there are at least 4 ways to do question answering in @LangChainAI? https://www.youtube.com/watch?v=DXmiJKrQIvg
Whose Opinions Do Language Models Reflect? “In fact, recent reinforcement learning-based HF models such as text-davinci-003 fail to model the subtleties of human opinions entirely – they tend to just express the dominant viewpoint of certain groups (e.g., >99% approval rating for Joe Biden).” https://arxiv.org/abs/2303.17548
Offline Reinforcement Learning with Reverse Model-based Imagination http://thu-iiis-ai.cn/en/archives/353
Extract the tools & technologies a company is using from their career page. https://twitter.com/GregKamradt/status/1643027796850253824
GPT-4 fails Steve Landsburg's economics exam https://www.thebigquestions.com/2023/04/05/gpt-4-fails-economics/
BanditPAM: Almost Linear Time k-Medoids Clustering via Multi-Armed Bandits https://news.ycombinator.com/item?id=35445312
Behind the curtain: what it feels like to work in AI right now https://robotic.substack.com/p/behind-the-curtain-ai
How I Learned to Stop Worrying and Love Nuclear Waste https://www.youtube.com/watch?v=jM-b5-uD6jU