Chinese open source AI releases, wipes the floor with US AI. Exposes AI as massively overvalued. Not surprising considering how big tech is just taking everything. Nvidia loses like half a trillion in value.
I’m happy to see a home run against the fascists that hurt them more than violence, at least to start. And I’m not a fan of AI (as hyped), but if it’s going to happen I’d rather it be with an open source performer as a base than the venture capitalists that abandoned us.
And really the US is no more trustworthy than China for those of us under the shadow of the big dick energy stuporpowers.
Not overvalued exactly. I believe the main point is they claimed to have used much much less power to train it, and that it demonstrates equivalent capability on much much less power.
There’s been a huge stock bubble on speculation of near infinite demand for datacenters, and NVidia processors. If DeepSeek’s claims are true, they just popped that bubble. We don’t suddenly need so many new datacenters. Demand for NVidia processors is not infinite, and their stock price is excessive compared to reality
The model has no censorship. Some websites running the model are hosted in China and thus do. Use one of the thousands of instances that run out of the US or run it yourself. It’s not like chatgpt or other openai trash, it’s system requirements are pretty low.
My horrible take only based on a few snippets:
Chinese open source AI releases, wipes the floor with US AI. Exposes AI as massively overvalued. Not surprising considering how big tech is just taking everything. Nvidia loses like half a trillion in value.
I’m happy to see a home run against the fascists that hurt them more than violence, at least to start. And I’m not a fan of AI (as hyped), but if it’s going to happen I’d rather it be with an open source performer as a base than the venture capitalists that abandoned us.
And really the US is no more trustworthy than China for those of us under the shadow of the big dick energy stuporpowers.
Happy to see free market working as intended.
Wait til they find a way to blame Biden for this!
Not overvalued exactly. I believe the main point is they claimed to have used much much less power to train it, and that it demonstrates equivalent capability on much much less power.
There’s been a huge stock bubble on speculation of near infinite demand for datacenters, and NVidia processors. If DeepSeek’s claims are true, they just popped that bubble. We don’t suddenly need so many new datacenters. Demand for NVidia processors is not infinite, and their stock price is excessive compared to reality
That open source performer is very much aligned with Winnie the poohs censorship, which should be worrying.
The model has no censorship. Some websites running the model are hosted in China and thus do. Use one of the thousands of instances that run out of the US or run it yourself. It’s not like chatgpt or other openai trash, it’s system requirements are pretty low.
Run one and ask it about Tiananmen. Post your results here.
They won’t do so of course, but I just compared llama (mistral 8b) vs deepseek r1 8b …
You won’t be shocked by the result. Mistral dutifully answers June 3-4 1989, violent crackdown, etc
Meanwhile WumaoSeek says “I’m sorry, I cannot answer that question. I am an AI assistant designed to provide harmless responses.”
One fascism vs. the other fascism, and everyone loses.