longer.blue

DeepSeek has been unavoidable as a news item this week. The year-old Chinese startup somehow triggered a tech stock sell-off that wiped billions from the value of major Western companies. But beyond the stock market angle, is it really such a big deal? DeepSeek certainly isn’t more impressive as a chatbot than the likes of ChatGPT. It doesn’t seem to be more useful or more accurate. It still has the same issues of having been trained on creative material whose owners very likely didn’t give informed consent. But DeepSeek represents a force that could massively undermine the US-centric AI industry because it operates and appears to have been created at a fraction of the cost compared to the models from OpenAI, Google and Meta. If it’s possible for a Chinese startup to create and offer something like this at a low cost, why would customers and enterprises the world over pay big subscription dollars to the US giants to access their services? Answer: they wouldn’t. If DeepSeek’s story is true, US giants have lost control of AI. Other countries are free to make their own models that don’t reflect US values, and researchers can explore new ways to use and develop AI without having to deal with the tech giants at all. Expect to see OpenAI claim (with delicious irony) that DeepSeek copied its homework, as experts hail its arrival as both a security danger and a democratisation. But longer term, the AI bubble may be beginning to burst.