Microsoft’s 3 billion bet on Copilot is assessed to be a disaster

[curl-6]

Norion said:

curl-6 said:

AI speeds up some processes, but it also makes mistakes, requiring additional time to be spent fixing these errors.

AI itself also cannot innovate because it cannot produce original ideas, it can only remix and regurgitate whatever data is fed into it.

It's a useful tool for some tasks, like say upscaling, but there's major limits on what it can reliably do, and it's mostly not cost effective as it's expensive to run at scale. AI companies are bleeding billions of dollars and are kept on life support by investors who still buy into the hype.

The thing I posted earlier shows this isn't the case. It used reasoning to disprove something mathematicians believed for close to a century and produced an original idea that has advanced the field of mathematics which is a huge deal. I really think you're not up to date on just how capable state of the art 2026 models are. There's a significant gap between them and the best stuff from a year ago and a big gap between them and the free stuff. They're absolutely capable of producing original ideas now so the idea they just remix and regurgitate their training idea is outdated. 

Finding patterns and drawing conclusions in numerical data isn't forming an original idea, it's data in, data out. There's no "reasoning" involved, the model isn't really thinking.