This morning, DeepSeek is latest shiny gadget that is exciting (or unnerving) everyone who follows artificial intelligence; if your 401(k) is big into tech stocks and U.S. companies riding the AI boom, you may wince if you check your portfolio this morning. Andrew Ross Sorkin said on CNBC this morning, “it is mind-blowing, and it is shaking this entire industry to its core… it’s already being called a major breakthrough, as the app shows its work, its reasoning. It answers users’ prompts… I spent basically all weekend playing with it. It is the number one app, I believe, on Apple right now. The product overtaking ChatGPT, today, on the app store.”
The Economist, attempting to lay out the significance of DeepSeek in layman’s terms:
In December DeepSeek published a new large language model (LLM), a form of AI that analyses and generates text. v3 was almost 700 gigabytes, far too large to run on anything but specialist hardware, and had 685bn parameters, the individual precepts that combine to form the model’s neural network. That made it bigger than anything previously released for free download. Llama 3.1, the flagship LLM of Meta, the parent of Facebook, which was released in July, has only 405bn parameters.
DeepSeek’s LLM is not only bigger than many of its Western counterparts—it is also better, matched only by the proprietary models at Google and OpenAI. Paul Gauthier, founder of Aider, an AI coding platform, ran the new DeepSeek model through his coding benchmark and found that it outclassed all its rivals except for o1 itself. Lmsys, a crowdsourced ranking of chatbots, puts it seventh, higher than any other open-source model and the highest produced by a firm other than Google or OpenAI (see chart).
Chinese AI is now so close in quality to its American rivals that the boss of OpenAI, Sam Altman, felt obliged to explain the narrowness of the gap. Shortly after DeepSeek released v3, he tweeted peevishly, “It is (relatively) easy to copy something that you know works. It is extremely hard to do something new, risky, and difficult when you don’t know if it will work.”
Last week, Trump’s major infrastructure announcement was “Stargate,” a joint venture between three large tech companies to invest as much as $500 billion into building out U.S. artificial-intelligence infrastructure over four years; the three major companies are Oracle, Open AI, and Softbank, a major Japanese multinational investment holding company headquartered in Tokyo. You may recall Elon Musk publicly scoffing at the announcement.
I think one big lesson of the past five years or so is that when a group of Chinese scientists get together and start experimenting with new technologies, only good things can happen, because sooner or later, all of their biggest innovations go viral!
The other big lesson of recent years is that when a Chinese company releases a new app, there’s absolutely no harm in everyone downloading it to their phones, because Chinese companies are legendarily respectful of user data and personal information!
Back in 2023, a Harvard University symposium concluded that China is setting itself up to be a kind of global arms dealer in the race to apply AI to government surveillance and control — the best friend of every autocratic regime on the planet.