I think it is pretty hopeless": DeepSeek proves OpenAI's previous dismissal of AI startups with only $10M funding wrong, reveling in success at a...

[Windows Central]

In 2023, OpenAI CEO Sam Altman dismissed competition from startup AI companies, especially those with limited financial resources in the range of up to $10 million. "I think it is pretty hopeless," he added.

I think it is pretty hopeless": DeepSeek proves OpenAI's previous dismissal of AI startups with only $10M funding wrong, reveling in success at a... : Read more

 

[GraniteStateColin]

If, per the Windows Central article about DeepSeek distilling Open AI's data (https://www.windowscentral.com/soft...ng-data-without-permission-fails-to-see-irony), DeepSeek did it so inexpensively by piggybacking on the work OpenAI already did, then their accomplishment isn't really an accomplishment. Interesting to watch and see what happens next.

 

[Dosmastr]

Not sure I agree.
They say they used flagship models to train it. So we still need those.

 

[fjtorres5591]

If, per the Windows Central article about DeepSeek distilling Open AI's data (https://www.windowscentral.com/soft...ng-data-without-permission-fails-to-see-irony), DeepSeek did it so inexpensively by piggybacking on the work OpenAI already did, then their accomplishment isn't really an accomplishment. Interesting to watch and see what happens next.

Agreed.
There is also a report from a semi-credible news source that the model is fast and cheap, but inaccurate to the tune of 83% failure rate. And it seems to lack the added "guardrails" against misinformation and hallucination that western models are required to have. They quoted an "investor source" who claimed accuracy doesn't matter if it is 30 times cheaper. Yeah, right.

Also, the comparison is apples to oranges. The thing is actually an SLM trained by an unidentified (chatgpt?) LLM. A better comparison world be Microsoft's ORKA which is still in the experimental stages despite doing algebra last year. Smacks of a geopolitical stunt to counter OpenAI $500B project..

Finally, it's still a chatbot. That is not where the real value of generative software lies.

At least it deflated the idiot investors' bubble.
Let them chase the new 3 day wonder chatbot while serious investors stick with industrial and enterprise grade products.