Tuesday, 28 January 2025

A Thing Called DeepSeek

That high pitched noise you heard yesterday was the sound of the worlds smallest violin playing for the Tech billionaires as almost £100bn was wiped from their stock thanks to a Chinese start up company called DeepSeek.
While American Ai companies spent hundreds of  billions developing their Ai, the Chinese created one which according to experts is just as good but they spent only £5.6m on it.
DeepSeek, an open-sourced problem-solving model whatever that is, uses far less computing power and far fewer chips, 50,000 against the 500,000 of OpenAi, therefore far less money, roughly 3-5% of the development costs for ChatGPT – to achieve the same or better results as its US counterparts.
Since DeepSeek was able to create such a product with a fraction of the resources other AI companies use, the major Chip manufacturer, Nvidia, saw hundreds of billions wiped from it's market value, the biggest drop in the US stock market’s history.
The company behind DeepSeek was only founded in 2023 and although the specifics of how it does what it does for a fraction of the cost is a whole world of geek speak but was likened to having a huge team but only calling in the experts you actually need for the task which doesn't really explain it but easier to understand than the 1.8 trillion parameters of OpenAi against the 671 billion of Deepseek.
No, me neither.

1 comment:

Not really a blog said...

DeepSeek used ChatGPT to obtain its abilities. Meaning, without ChatGPT, there would be no DeepSeek. An obvious move, but investors don't understand IT, and IT experts don't understand economics.

I don't know anybody that predicted this, but I do know some people that aren't surprised by this... Apparently the AI development companies need to run some business war games and competitive scenario analysis.