This is John Batchelor, conversation with a colleague Jack Burnham of the Foundation for Defensive Democracies,

about the contest between the United States and Europe and the People's Republic of China,

pursuing better and better AI functionality, efficiency, productivity, trustworthiness, building toward the future.

And on the basis of the microchips, that is possible.

But China is banning its major Internet firms from using the best microchips now available.

That would be NVIDIA.

In the course of reporting on that, Jack comments on DeepSeek that gained a lot of attention when it was first introduced as cheaper, faster, just as good.

The questions I learn remain.

Here's Jack to explain.

Deep Seek, question mark, and Deep Seek working to the future much more of this later deep seek really became

revolutionary in part because of its low costs and its open source structure so rather than

having a proprietary model something like chat GPT, for example,

or Claude, you know, DeepSeek was an open source model, so you could download it onto your

computer and run it locally in a way that you really couldn't with other types of large-scale

AI models. So that open source combined with, it's overall relatively high quality, particularly

around things like mathematical computation, which are, at least initially, were types of

problems that other AI models like chat GPT struggle to solve, really made it quite appealing.

Now, you know, according to recent studies, particularly from the

National Institute of Standards and Technologies, they've found that Deepseek maybe doesn't perform

quite as well as other types of both open source and proprietary models that have been

generated by American companies in particular.

But at the same time, you know, DeepSeek is what I would consider a good enough AI model.

...