Right Now Sure Looks Like the AI Peak

First, a concise summary of some things I've argued in the past:

  • "AI" is algorithms running on computer hardware.
  • We actually have a pretty good concept of the capabilities, and the limitations, of algorithms and the computer hardware they run on. Specifically, we know there are certain classes of problems they are very good at solving, and others they are unreliable or even terrible at.
  • There is nothing magical about any existing or potential future "AI" that allows it to transcend these limitations.

This leads to some business conclusions:

  • There are some problems that are well-suited to being solved by algorithms; we mostly have custom-built, special-purpose algorithms already out there solving these problems with a high degree of reliability and scaling, and relatively low cost.
  • There are some problems out there that are ill-suited to being solved by algorithms.
  • There are some problems out there that are amenable to being solved by algorithms, but are not currently being solved by algorithms. Any genuine value created by "AI" can only come from identifying and solving these problems. Any net value that can be pocketed by corporations/investors will come from here, minus the (often substantial!) costs of creating and running the models.
  • The only way AI makes a lot of money is if there's a lot of that class of problem, and a lot of value in solving it. There's not.

So with that established, the investment case for AI as a technology is looking rather shaky. I've been saying this for a while, but there's a big obvious question following along here:

When will the market figure out that AI is a boondoggle?

There's an old saying: "the market can stay irrational longer than you can stay solvent". But the market can't stay irrational in this specific way forever; eventually it will have to become irrational in a completely different way. Specifically, just about everyone - from retail investors to VCs to corporate behemoths - who's put money into AI has done so because they expect to get a lot more money out of it, soon. Retail investors, VCs, and corporate behemoths are not known for their patience or their willingness to give away money forever! So while it's tough to call when a particular run of irrational exuberance will exhaust itself, we can guarantee that it will happen eventually.

Nevertheless, I think there are signs. It seems significant that Meta just pissed off a bunch of investors by talking too much about AI on an earnings call. It seems especially significant that the specific thing that made the investors upset was an admission that they were spending a lot of money on AI, and that they didn't expect to see major profits from it for a few years.

It also seems significant that no one has come out with a valuable customer-facing AI application. Stuff like Midjourney and ChatGPT is nice, sure, if you need to generate a throwaway image or some throwaway text and you're willing to compromise on quality. But as far as I can tell, the big corporate AI applications that keep getting advertised on TV and all mostly involve bundling AI into existing tech-company offerings in hopes of making them marginally more valuable - oh look, GPT autocomplete in Microsoft Word! Oh look, JIRA lets you search and summarize with a language model that reads your tickets! Oh look, IBM has autocomplete for your code! Oh look, Google has a (rather controversial) image generator! And not only do all these products kind of look the same (how many image generators can possibly be profitable, even if some of them are extra-woke and some of them are extra-not-woke and some of them let you generate porn and some of them lecture you if you ask?) none of them are apparently even worth selling on their own instead of as add-ons to existing apps.

Meanwhile, attempts to actually use AI to perform tasks in the real world have most been catastrophic, expensive failures. That's another revenue source that the hype never delivered on!

So overall, I think we're close. I can only hope the crash is messy and comprehensive, and that it only harms those who most richly deserve it, and leaves the innocent better off.

You'll only receive email when they publish something new.

More from Tom
All posts