If DeepSeek is open source and found a new way to train low cost…
Doesn’t that just mean other AI companies will fork and use that new way and multiply it against their massive compute power?
Someone tell me how this negatively impacts them? If anything, it’s a positive bc less spend needed going forward
agreed. It is a massive win. Just an excuse for traders to take some profits. This is just a short news cycle.
this was my reaction
True but that's for "future Ai chatbots etc".
The one's that exists like by OpenAi is running and training on way too high cost..
The crash is about significant cost diff in US based Ai Vs China based...
The Deepseek phenomenon is really interesting.
Its a net positive for all. The reason the stock tanks has to do with expectations and that people arent rational when coming to investing.
The revealed almost everything they did under MIT Licences. I saw an independent estimate saying it is possible to train R1 for $2.5 million with opitmized Nvidia H100.
https://github.com/deepseek-ai/DeepSeek-R1/blob/main/DeepSeek_R1.pdf
This is the video where I found the information. If $LEO pumps enough, we could have our own R1 Soon™ 🤖
https://inleo.io/threads/view/vimukthi/re-leothreads-2safdezdx?referral=vimukthi