R to @mikeyoung44: doesn't matter for the headlines🤣

R to @GregBernhardt4: Yep that's what the people posting this want you to think

R to @mikeyoung44: ah, I thought these were copy and paste from the model to fake human writing

R to @GregBernhardt4: Why? These are quotes and examples of AI generations that are cited in the articles.

R to @williamlegate: They're quotes, the papers weren't written by AI and passed off as research like chaz is trying to imply

R to @DataChaz: Now actually read them, and you can see that these are quotes

R to @itsandrewgao: well you see it's the second iteration of a model fine-tuned on 70b params from chats between users who frequent huggingface to talk about NSFW ai-based game generation tools using screenshots of TI-83 programs

Researchers build a navigation model for robots that combines the way heat moves around insulators with a custom diffusion model https://open.substack.com/pub/aimodels/p/combing-thermodynamics-and-diffusion?r=2apyaf&utm_campaign=post&utm_medium=web

Well I think this is bad news for some agent-based frameworks that rely on "critiquing" agents like the one I reviewed yesterday https://notes.aimodels.fyi/gpt-4-doesnt-know-its-wrong-an-analysis-of-iterative-prompting-for-reasoning-problems/

RT by @mikeyoung44: On this day 1805 the Royal Navy defeated the combined fleets of the the French and Spanish Navies at Trafalgar

Well I think this is bad news for some agent-based frameworks that rely on "critiquing" agents like the one I reviewed yesterday https://open.substack.com/pub/aimodels/p/gpt-4-doesnt-know-its-wrong-an-analysis?r=2apyaf&utm_campaign=post&utm_medium=web

R to @batwood011: LOL

R to @mikeyoung44: Sorry we are getting rate limited :(

R to @mattshumer_: Didn't work at all?

New paper: 3D-GPT employs LLMs as a multi-agent system with three collaborative agents for procedural 3D generation Some of these look really good. Let's see how it works! https://aimodels.substack.com/p/3d-gpt-a-new-method-for-procedural

The amount of AI and ML research generated daily is growing exponentially For $60/year, I’ll read and summarize the top 5 papers per week for you So you never miss a breakthrough https://aimodels.substack.com/subscribe

R to @AIatMeta: "The findings also raise some important ethical considerations around privacy of decoded mental content..." Can't believe I just wrote that! https://aimodels.substack.com/p/decoding-visual-perception-in-the

R to @mikeyoung44: I actually read this paper as pessimistic for the current models tested

R to @emollick: Not sure that's the full takeaway here, the paper says that at best it may get close to the threshold to passing: https://aimodels.substack.com/p/can-gpt-models-be-financial-analysts