I've just seen this graph of the Cerebras CS-3 and I'm wondering what the catch is here. Is it really that powerful?

Automating AI research is exciting! But can LLMs actually produce novel, expert-level research ideas? After a year-long study, we obtained the first statistically significant conclusion: LLM-generated ideas are more novel than ideas written by expert human researchers.

Yesterday, Apple made a smart move by rebranding the concept of "Artificial Intelligence" as "Personal Intelligence." I think it was a brilliant rewording.

It is a shame that what was probably an error in the beginning ended with the reputation of everyone involved.

🔴 ¡Crea Podcasts de TUS DOCUMENTOS! La nueva app NotebookLM de Google es muuy potente! Te permite subir uno o varios documentos y obtener resúmenes, hacer preguntas o incluso generar una conversación entre dos IAs a modo de podcast sobre tus textos! 👇

Yeah. Anyone that is seeing the trend should be, at least, prepared. I use cursor. It is so cheap.

Oh my... this, with Claude.ai projects, would be super powerful

Uhmm this is an interesting take. So it seems true what some people said yesterday, o1 and GPT-4 apparently are not that different, they are when you find the correct problem for o1:

What a day it was yesterday. Google’s NotebookLLM is just amazing, then OpenAI new model explained as a podcast by NotebookLLM is just beautiful and so useful. What a time to be alive.

No surprises. It would be very easy to replicate a model if the CoT is returned! The moat here is "just" the dataset. Again.

This is what happens when user hardware is so close to professional hardware. I have 7 4090 and planning to sold them all if / when the 5090 is out (IFF the VRAM is > 24GB)

Someone said we are converging to a world where all the data will be AI-generated. I don't know if it is going to be that extreme, but definitely a huge percentage of the total will be AI-generated.

Didn't Leopold Aschenbrenner say that exactly this will happen?

Microsoft Search Copilot is just useless. I can't believe how fast they killed what was once a good product.

"AI companies don't make money" Tell that to the Stripe data. Top AI companies hit $30M in revenue 5x faster than their traditional SaaS counterparts.

This is a phenomenal work, but why have they decided not to share the model?