AI is the Ultimate Hold-My-Beer Event, et al.

A brief but definitely definitive word on AI

Is it weird that I haven’t written anything about AI?¹ I don’t know.

At one point, I considered writing a sublimely contrarian note promising to never write anything about artificial intelligence, which would be sincere in the sense that I had concerns about this technology, while a joke in the sense that no one asked for my opinion. At the same time, my recusal would be a practical decision, because AI is a very news-y subject and this isn’t a very news-y newsletter.² Finally, by refusing to step into the bottomless morass of popular discourse, I would be doing you, the reader, a subtle but profound favor: you wouldn’t have to read another take on AI.

I never wrote that note. So I guess my integrity, while always in doubt, at least remains unchanged.

---

Still the main reason I haven’t written anything about AI is that, like so many other things, I have no unique thoughts, and in fact don’t really know what to think. Given the range of responses over the past few years, I suspect no one else does either. Among other things, AI has been called magic, garbage, a miracle, a nightmare, thrilling, boring, liberating, colonizing, a world-historic catalyst for productivity, a massive waste of time, snake oil, a snake eating its own tail, a gift to be used on behalf of civilization (i.e., The Boromir Option), a curse to be rejected for the preservation of civilization (i.e., The Benedict Option), apocalyptic (i.e., an augur of the End Times), apocalyptic (i.e., a Revelation of what is and is not of ultimate import), et al.

All of the above are probably valid to varying degrees, but since we’re here, I’ll offer my own metaphor as a sort of thought experiment.

AI is the ultimate hold-my-beer event.³

What I mean is, when a guy lumbers up from his place by the campfire to announce “hold my beer,” he’s not saying “I can do whatever it is you think can’t be done.” He’s saying, “fuck it, I’ll give it a shot,” or “fuck it, why not?” The second clause will vary, but the first is constant. In more philosophical parlance, his declaration translates to something like, “I accept your challenge on the following conditions: we agree that it appears impossible, ergo I can’t be held accountable for the outcome.”

Three propositions follow from the above premise:

1. Our would-be hero has only imagined a challenge. In reality, no one asked him to perform the task.

   • The hero thinks he is Prometheus; his audience thinks he’s Don Quixote.

2. Because the task appears impossible, or even pointless, whether our hero succeeds or not doesn’t matter. What matters is the attempt.

   • In the same breath that he says “hold my beer” he also says “the action is the juice.”

3. Because our hero initiates the endeavor, and because its outcome is moot, he controls the narrative. He alone gets to determine what counts as success.

   • Don Quixote may be a fool — but he is a clever, entertaining fool. If the task appears impossible, then all he has to do is appear to accomplish it. If he can cross the threshold of plausibility, then the joke is on us for doubting him.

This is all very underdeveloped and overdetermined, but hopefully the general train of logic makes sense. The way I interpret our hold-my-beer hero is the same way I interpret the state of AI: sensational, audacious, occasionally impressive, but with neither a clear purpose nor any clear metrics of success.

---

How much of this is a problem? That’s also unclear.⁴

For AI evangelists all problems are ancillary, temporary, or illusory. Ethical problems can be solved by the claim that AI’s potential benefits vastly outweigh its potential hazards. In the meantime, technical problems will be solved through technical solutions. For example, any risks AI may pose to climate change due to its enormous energy requirements are negated by the argument (read: assumption) that AI will solve climate change.⁵ Finally the claim that AI is inevitable serves as a kind of trump card. If AI is the permanent and prevailing power of the future — and the future is now — there’s nothing to be done. In fact the tables have turned.⁶ Skepticism, much less resistance, is futile in the technical sense and dubious in the ethical one.⁷

But to return to the hold-my-beer analogy, the main problem is much more straightforward and immediate: AI can’t be held accountable. In fact, because there is no self-evident way to evaluate its performance or even its objectives — no standard, criteria, or precedent — AI is beyond judgment. For the hold-my-beer hero, this is a feature, not a bug. Yes, he may be bored. Yes, he is almost certainly drunk, and yes, contrary to what I suggested earlier, a few powerful people (also drunk) actually did ask him to play Prometheus.⁸

But mostly he attempts his task because fuck it. If there are no metrics of success, then there are no stakes. And if there are no stakes, then the outcome is rendered meaningless. The hero can do whatever he wants. The game is an exhibition, and all that counts is the spectacle, in which, at the risk of sounding dour, “everything is permitted,” as Ivan Karamazov said.

1. I am using the catch-all term “AI” out of convenience. Obviously there are a wide variety of AI technologies, and the term is overused. The FTC describes “AI” as a marketing term. I think most of what I have to say here can be applied to AI writ-large, including chatbots as well as the soon-to-come “AI agents.”

2. What’s true about AI one day may be less true, or not true at all, the next day.

3. But wait! What is an event?

Philosophers seem to have complicated this otherwise straightforward question, as evidenced by the Stanford Encyclopedia of Philosophy’s entry.

For our humdrum purposes, let’s just say an event is a thing that happens. Some events are clear and concrete — concerts, parades, athletic competitions, etc. — while other events take on a figurative aspect, for example, when we say Ken Bone or the Internet of Things “is having a moment.” More recently, and at much larger scale, we could say the Eras Tour or “Barbenheimer” qualified as events on both literal and figurative levels, and in doing so achieved the status of “phenomenon,” which — again, for our non-Kantian purposes — we’ll colloquially define as uniquely remarkable events.

Still, the character of an event dictates that it will come and go. Some events become completely integrated into ordinary life, therefore ceasing to be events (e.g., the Internet of Things, which is now just…stuff you expect to be wi-fi enabled), while other events cease to be part of ordinary life but like embers that still flare and glow in the aftermath of a fire live on in our memory as events (e.g., Ken Bone, to whom our nation may yet turn its lonely eyes).

For better or worse, AI will probably become integrated into ordinary life. But for the time being, AI remains primarily an event — and is being advertised as such.

4. For what it’s worth, I’m not sure it’s a problem that a general-purpose AI tool like ChatGPT lacks a clear purpose, or is “unscoped,” as Emily Bender has argued. Good tools often leave space to experiment in a way that’s not strictly instrumental or utilitarian, which explains why, from a distance, we admire our hold-my-beer hero. He is unbidden and unbothered, if a little tipsy, but so it goes with human freedom. He celebrates himself and sings himself!

5. Big, if true!

6. Or, to borrow again from a philosophy I only superficially understand and am possibly misappropriating, a “transvaluation of values” has occurred.

7. Obviously this ethos opens the way for the sort of AI doomer scenarios. To be clear, I don’t think those arguments are persuasive. As others have pointed out, companies like OpenAI benefit from dystopian fears by positioning themselves as the only available bulwark, which means they get to make the rules (a go-to playbook move in Silicon Valley).

8. The powerful few are powerful, indeed! But they are not the happy few. Not really, anyway.