DeepSeek, Deep hypocrisy

FromTheFields Tuesday February 18, 2025

The weekend before last ,a Chinese hedge fund, High Flyer Capital Management, released a new Artificial Intelligence large language model, DeepSeek-R-1 which appears to be equivalently functional to Silicon Valley created ChatGBT. And they claim it only cost $6 million, a fraction of the cost of its Silicon Valley competitors. Moreover, it's claimed that it was built with older pre-embargo Nvidia chips.

Of course, I decided to put it to the test. Predictably, the site was incredibly busy. It took me a couple of days to establish an account. Then I asked my standard test question for AI: "Would modern civilization be possible without the use of fossil fuel?" It gave the samed hedged answer that I've received from the U.S. based AI sites, "Yes, but history would have looked much different...yada, yada, yada." I tried to ask it to give me a history of Tiananmen Square but it was too busy to respond. Hmmm.

OpenAI is already complaining about copyright infringement. Talk about the pot calling the kettle black. The real lesson to be learned is a demonstration of the law of unintended consequences. The U.S. has banned export of Nvidia's latest generation high end chips to China. Can't have the ChiComs using any of our vaunted technology, right? So the clever Chinese evidently found a low cost way to produce a product that appears scary enough to spook the stock-trading algorithms to tank Nvidia stock 17% in one day.

Ultimately, Nvidia probably has nothing to be afraid of. More companies producing more AI machines will just feed an insatiable demand for machine written term papers and the like. And there may be some actual productivity gains coming from AI in the real world. It sure reduces the research time to write these Reports. And that's this week's Report From the Fields. I'm Richard Fields. See you again next week.


Permalink: https://libertariancounterpoint.com/tiki-view_blog_post.php?postId=134