Sam Altman Says "Ai Progress Will Be IMMENSE From here, Llam 3.1 BEATEN Already! Google news..

Says the company who just appointed a US military general to their board. 🙄

immense maybe for them, but not for us.. we don't even have the new voice engine yet

Tired of the promises and hype, Sam. We are all learning to live without OpenAI.

Good luck ClosedAI beating Llama3, because it's not enough to be a "little better". It will have to be immensely better to compensate being closed source.

Owner hypes up product ?!?! noooo wayyyyyyy

Sigh….. but I’m still working

Take a moment to appreciate being part of the workforce of human labor. Humans being essential to the economy will be a part of history soon.

XAi did not complete a 100 GPU cluster in 19 days. You can see that he almost said completed but caught himself. I would estimate that 30K maximum are in the cluster currently. I hate having to parse everything Elon's says but history shows he exaggerates and inflates everything.

I can't believe that I can run a 4-bit quantized version of Llama 8B completely offline on my Galaxy S24 Ultra. It's slow, but definitely useable, and pretty smart. Definetly better than GPT 3.5 . We've come so far in such a short amount of time. I don't know much about tech and had to ask perplexity how to set it up. I wonder if we'll ever get huge context window models running locally on phones? Not sure how it works or if that's possible, but context window size is underrated. It's the only reason I ever use Gemini. I'll upload like 20 big PDFs at once.

AlphaProof might seem unremarkable to ordinary people, but to AI practitioners, it's an absolutely groundbreaking breakthrough.

OpenAI is working SOOOOOO hard on safety. But what's the point if your company's products fall behind Open Source? No one will use your models and so your company will die soon into history. As Meta plans to release Llama 4 by the end of this year, If OpenAI does not release GPT-5 and all their promised products soon, 2025 is the year OpenAI falls apart.

Are we sure this search AI isn't just Google search from circa 2007 back when it actually worked really well with a summarizer on top?

I am very pleased with you highlighting the importance of the AI breakthrough in solving math problems at IMO level. Great channel, I regularly watch your videos!

Let's see, let's see. So far, everything is going very slowly. My grandparents also watched movies where robots serve people and they have not yet appeared in our apartments as servants.

Democratic AI to complement democratic bombs 👍

"Llama 3.1 beaten already" Llama was at no point ahead, its just the best open source model

The three uses for AI agents that are always talked about are customer service, flight booking and email answering. I hope we can do better than that. That sounds really boring.

From the things Sam Altmann said about the international implications of AI leadership, it definitly opened up a new perspective for me on that topic, as i definitly do agree on the points he was making. It's not an AI race, it's an AI war. However what he didn't mention and what is the conclusion that i came up with myself is that china will have a very difficult time taking the lead in AI if it's ever taken in the first place, simply because of nvidia being an american company and by far the most relevant provider for increases in ai training computing power, which i don't see changing ANYTIME soon AT ALL. it's not even close by any means in any way anywhere. So in that very relevant aspect, China will likely fall behind a lot, if it ever gets close to overtaking the US in AI development since the us government will likely intervene through nvidia at that point.

Yeah, Sam Altman is talking about ensuring "democratic" lead in AI over authoritarianism, while being closed source and putting the previous NSA director on the board together with a former Fed chief. What's his suggestion for "democracy"? Opening up AI for the greatest number of people? Or petitioning the US government to make AI research illegal for anyone but OpenAI and a couple of other US-based mega-corporations? Likely the latter! That's not authoritarian at all! No! That's very democratic!

First tell me why we can't select between 4O or Mini Model. Instead of only regenerate with gpt mini