lol... dude. It's already amazing at coding. It can solve math problems better than most engineers or people with a bs in math. It writes better than 99% of the population. there are a limited number of things that AI does worse than humans, but for the rest, it's a fucking genius.

I have an s24 ultra and initially i thought it would be something useful... not so much. I use AI all day, every day (around $300-500 worth of claude code credits per day) plus grok, openai, and Gemini. The local models arent' really all that useful right now.

they have destroyed some of the plants.

It's really easy to tell them what it is training on- Everything!

Softbank doesn't have nearly enough money to fund this and probably can't get enough investors or loans for a trillion dollars.

What is the electricity cost per kw/h in the UK?

Let's scale up the business by selling more cars that cost more to make they we can sell them for... Good plan! https://ancillary-proxy.atarimworker.io?url=https%3A%2F%2Fwww.motor1.com%2Fnews%2F71...

Failure is an essential part of capitalism. Everyone will be worse off if a single failure eliminates you from the market.

If only there was a company with thousands of communications satellites already in orbit that could handle network traffic...

This won't end well for California.

They are not rising any more. Just the opposite.

solving math problems that Fields prize mathematicians struggle with is crappy? Your criticisms are based on really old data. You really should keep up. The models are very good now.

No, they could not. The new supercomputers are using the latest Nvidia GPU's and they can't be sold to China, and are already in short supply. (even before Elon builds his 1 million GPU behemoth.) China's latest model is great for opensource, but it is still not leading the world. OpenAI is in the lead with O3 and I would expect xAI to pass them soon, since they have an overwhelming compute advantage, which is only going to increase.

I'm hoping a 72B model is released soon.

define good. You can run a pretty damn good model on 2 3090's or 2 P40's (total 48 GB vram each.) Apple devices could also do it well if Apple would reduce the memory prices. Give it another generation of hardware (but not from Nvidia, they are intentionally trying to keep inference costs high.)