This week, the Chinese company DeepSeek sent shockwaves through the AI industry when it released a new, free chatbot that seemed to rival the performance of OpenAI’s flagship GPT-o1.
DeepSeek claims that they trained their DeepSeek Chatbot for only $6 million — a tiny fraction of the billions spent by OpenAI and its other American rivals.
Markets promptly freaked out. In a single day, Nvidia, which supplies high-end computer chips to the AI industry, lost over $580 billion in market value. That’s the largest drop in value in Wall Street history.
Commentators called DeepSeek “an extinction-level event” for the venture capital firms funding America’s wildly expensive AI systems.
As an AI expert, I’ve extensively tested the new DeepSeek chatbot.
I’m not impressed, and I expect that DeepSeek will be more of a tiny blip than a catastrophic event for the AI space.
Here’s why.
The Sincerest Form of Flattery
Imagine that you want to create a brand-new search engine. The ideal way to do this is to crawl every webpage on the internet, assembling a massive database of text, links, images, and videos.
You’ll then want to develop powerful algorithms to sort through all this data, housing them on massive servers in sprawling data centers.
You’ll also need a source of user data to gather info showing what people engage with on the internet and how they respond to the content you’ve gathered. You’ll probably need to buy a big social media platform (I hear TikTok is for sale…)
The whole process will cost you about $20 billion.
Of course, there’s an easier way.
Instead of building your own infrastructure for crawling and processing the entire internet, you could…
Leave a Reply