DeepSeek and the Satellite of Love

DeepSeek is a Chinese artificial intelligence company. It made big news over the weekend by announcing that it had built AI models which are more cost effective than anything any American firm has come up with, and on top of that, it's made them open source. (Open source here means that DeepSeek has apparently made the underlying code available to users.)

Silicon Valley and Wall Street went berserk today as a result. Nvidia is the company which makes a lot of the computer chips which drive AI, and Nvidia's shares dropped by over 16%. Alphabet (Google) stock dropped by over 8%. Microsoft stock dropped by over 9%. For reference, Microsoft's market cap is about $3.3 trillion - that's $3,300,000,000,000 - so a 9% drop is about $300 billion. That... is a lot of money. (Interestingly, stock prices for Meta (Facebook), Apple, and Amazon all went up a little today.)

In terms of AI models, the biggest players in the U.S. are Google, Amazon, Microsoft, and Meta. Apple is a little late to the game but they're in play too. And you may have heard of newer companies like OpenAI and Anthropic. Now, I'm just listing them here. I have no particualar opinions about any of their offerings relative to one another. For a number of reasons, AI has not been something that I have found especially interesting.

It may seem surprising for someone who "works in tech" to express a lack of interest in AI, and I want to elaborate on that. There's the hypothetical potential of AI, and then what it seems like these companies are going to do with it. I think that in hypothetical terms that it's extremely unclear what all AI could do, but somewhat likely that it will be brought to bear to do a lot of interesting and maybe even wonderful things, perhaps in fields like medicine. But that's mostly hypothetical. If instead we talk about how AI is being positioned and what these companies seem to be trying to do within the market, I am not impressed. Indeed, I think they're mostly full of shit, if not worse. Matt Stoller ripped into American tech last night, and this quote is representative:

More broadly, as with Boeing, Intel, General Electric, et al, the incentives in the American corporate world are to win by innovation around monopolization and extractive techniques, not by building more and better products. Killing the underlying operational capacity of one’s firm always looks good, as you generate more and more cash. Until, as with Boeing, it doesn’t.

Boeing is the best current example of a huge firm gone slothful to the point where their actual products have suffered, and since their products are airplanes, this means that their corporate morass has literal life-and-death consequences. Believe me: there's about no harsher insult you can lob at Microsoft than to compare them to Boeing.

To put it another way: I don't believe that these companies are interested in building great products that benefit consumers. They're not necessarily opposed to true innovation, but that's of marginal importance. What matters is "winning", defined however they please at any given point in time. That might mean pure profitability, it might mean market share, it might mean market cap, it might mean some crazy formula that slices and dices those numbers. And while I don't think I'm talking about anything particularly novel - corprorations like money, big news flash there - I do think that American tech in general used to be at least a little different. There used to at least be pride in the product... yes, I think, even at Microsoft.

One of the hot takes today has been to compare DeepSeek to Sputnik. This is actually the thing which really got my head spinning and made me want to write a bit, because the Sputnik comparison makes a little sense superficially, until you actually really think about how absurd and potentially dangerous it is.

Sputnik was the first satellite, launched in 1957 by the Soviets, who, in the process, beat the Americans to space. People freaked out. Within American power circles, this was unacceptable, it was a matter of national security, blah blah blah. Among other things, what came of it all was John F. Kennedy's space program. In a May 25, 1961 address to Congress, Kennedy laid out a vision to put a man on the moon within the next 10 years. The American government broadly committed to supercharging the American aerospace sector. (It helped, of course, that the aerospace sector was substantially the same as the defense sector.) The thing was that in 1961, Kennedy's statement was something which could galvanize the nation against a common external threat and behind a common internal quest. But what could possibly galvanize American sentiment today behind anything at all? (Maybe the price of eggs... but that's a subject for another time.)

And so Sputnik seems like a parallel for DeepSeek today. America has one obvious major international rival in China. China just beat America at something America shouldn't be losing at. America should take stock and come up with a plan to make sure that this was a blip. We should pour even more money into even more AI research to make sure we stay the world leader.

There are a couple of problems here though, not the least of which is this point Stoller made:

That said, no one denies that DeepSeek’s technical accomplishments are impressive, and the accomplishment suggest that there’s a lot about AI that we still don’t know. And a Chinese firm releasing a better cheaper product they claim was produced for less than most top AI project managers make in a year does make me feel like America is quite… Soviet.

See: DeepSeek may have caught American tech by surprise, but it's not because the American firms weren't already pouring themselves into AI. Sputnik won by being first. DeepSeek - and therefore the Chinese - just scored a win, though, not by being first, but by being better.

As is, the tech firms are already trying to shove AI down the collective throats of Americans. They're already trying to bring many more energy hog data centers on line, which means bringing more energy sources online. They're even trying to reopen one of the Three Mile Island reactors! The American people however aren't being consulted about all this. We're not hearing why all of this is needed, why all of this energy is needed, why we're likely going to have to pay more for water and power in many places, why we're going to be headed the opposite direction in meeting clean energy standards. We're being told all of this is necessary, but we're not being told why. And that's because they don't have to tell us! Companies building huge data centers have done so under fake names, and have forced local governments into agreements which are being kept secret from people. If all of this was being done for the good of Americans, then why does it all have to go down like this?

There are two problems, and ultimately, I think they’re the same problem. First, American tech firms have just been exposed by a rival. Second, even though the American people are having AI shoved down their throats, they’re not really very supportive of it. And the common thread here is that American tech isn’t pursuing AI because of any intrinsic or potential good. They’re not trying to make something better, they’re solely trying to make something more monetizable, and that means figuring out how to force people to see it and use it, instead of figuring out how to make it good. Indeed, the number one thing a lot of people hear about AI is how it can replace certain service type jobs - as in, actual jobs being performed by actual people who need to make money to get by. Oh, and, the AI isn’t actually cheaper than the workers, by the way. AI is extremely expensive!

Now, I'm not going to argue that it's preferable for the Chinese to do AI better than the Americans. But the American firms involved here are all monsters. And if they manage to spin DeepSeek into a "Sputnik moment", we're all going to suffer for it.

I have not tended to regard myself as a "tech leader". There are a number of reasons for this, primary among them being that I'm not often an early adopter of bleeding edge technology. My degree is in history, after all, and so I tend to like backward instead of forward, which, you know, maybe isn't how "tech leaders" usually think.

I’ve been rethinking this “tech leader” thing lately though. If we’re in the midst of a “Sputnik moment”, I think we may have real lessons to learn, but are in danger of having that spun wildly out of control. Maybe we need more leaders in tech who can actually think critically instead of just wildly screeching at each other that we need more AI so we don’t fall behind.

None of this should suggest that I am, or am about to become, a prime authority on AI. That’s okay. I cede that space to others. There are already a lot of examples of positive contributions AI has made. Those examples are the directions which I think should help guide further exploration in the space. But I know a thing or two about history, and I know we absolutely do not want AI to become the next frontier of a costly arms race. Space was that frontier once, and as it so happened, the United States did ultimately win that race, not by sprinting, but by enduring. We’re not a country prepared for that kind of marathon right now though. China is. We have to be smarter, which means we need political leadership which understands some of this, and we also need technical leadership which understands that financialization is not true innovation.

This may well be a “Sputnik moment”. But DeepSeek may not be Sputnik. American tech might be. And that is a very frightening proposition.