The recent selloff in US tech giants like Nvidia and Broadcom has been fueled by panic about a new Chinese AI startup, DeepSeek. The company claims to have made significant breakthroughs with its AI models that can perform on par with advanced US-born technology at a fraction of the cost.
DeepSeek's latest model was trained for $5.6 million in computing costs, which is significantly less than what it takes to train other state-of-the-art AI models. For comparison, OpenAI's GPT-4 model, launched last year, cost over $100 million to train, and Anthropic CEO Dario Amodei recently estimated that training some models can reach as high as $1 billion.
This news has sparked fears about the competitiveness of US tech companies in the AI space. However, it's worth noting that DeepSeek's achievement may be exaggerated by the hype surrounding it.
While DeepSeek's breakthroughs are impressive, it's unclear how scalable and sustainable its model training costs will be. Additionally, there's no guarantee that the company can maintain its current cost advantage as it expands its operations.
It's also worth considering that the AI landscape is constantly evolving, and what works today may not work tomorrow. It's possible that other companies will soon catch up or surpass DeepSeek's achievements in the near future.
In any case, this news highlights the rapidly changing dynamics of the AI space and the need for US tech companies to stay competitive and innovative to remain relevant.
DeepSeek's latest model was trained for $5.6 million in computing costs, which is significantly less than what it takes to train other state-of-the-art AI models. For comparison, OpenAI's GPT-4 model, launched last year, cost over $100 million to train, and Anthropic CEO Dario Amodei recently estimated that training some models can reach as high as $1 billion.
This news has sparked fears about the competitiveness of US tech companies in the AI space. However, it's worth noting that DeepSeek's achievement may be exaggerated by the hype surrounding it.
While DeepSeek's breakthroughs are impressive, it's unclear how scalable and sustainable its model training costs will be. Additionally, there's no guarantee that the company can maintain its current cost advantage as it expands its operations.
It's also worth considering that the AI landscape is constantly evolving, and what works today may not work tomorrow. It's possible that other companies will soon catch up or surpass DeepSeek's achievements in the near future.
In any case, this news highlights the rapidly changing dynamics of the AI space and the need for US tech companies to stay competitive and innovative to remain relevant.
I'm not too sure about all this hype around DeepSeek... $5.6 million for a model that's supposedly on par with US-born tech? That sounds like a pretty big claim 
. And what's the deal with these huge training costs anyway? It's like they're saying "we can make AI magic happen, but it's going to cost us an arm and a leg 
". Meanwhile, I'm over here just trying to get my email inbox under control... 
like whats next? china gonna steal all our jobs with their fancy ai and we're just gonna sit here and watch ? 
and whats with all the hype around it anyway? is it even that revolutionary ? 
theyre claiming to be able to train ai models 4 pennies on the dollar but is that really possible? i mean sure its awesome but what if its just a fancy trick? and whats gonna happen when they scale up their operations? will they still be able to keep it affordable? idk but im definitely keeping an eye on this one 
... just another example of investors getting cold feet about innovation