I recently finished a philosophical science-fiction novel by Soviet authors Arkady Strugatsky and Boris Strugatsky called “Roadside Picnic”. Aliens visit certain places on Earth; those places (“the zones”) become riddled with alien artifacts that are incomprehensible to humans, and an entire subculture and black market develop around discovering and re-selling those artifacts. As an example, they found eternal accumulators (called “so-so” or “«этак»” in the original) that people would procure to power their cars and houses. Yet, they didn’t understand how they worked.

One of the scientist characters in the book suggests that the alien visit can be understood as a sort of a “roadside picnic”:

But what about the Visitation? What do you think about the Visitation?”

“My pleasure. Imagine a picnic.”

Noonan shuddered.

“What did you say?”

“A picnic. Picture a forest, a country road, a meadow. A car drives off the country road into the meadow, a group of young people get out of the car carrying bottles, baskets of food, transistor radios, and cameras. They light fires, pitch tents, turn on the music. In the morning they leave. The animals, birds, and insects that watched in horror through the long night creep out from their hiding places. And what do they see? Gas and oil spilled on the grass. Old spark plugs and old filters strewn around. Rags, burnt-out bulbs, and a monkey wrench left behind. Oil slicks on the pond. And of course, the usual mess—apple cores, candy wrappers, charred remains of the campfire, cans, bottles, somebody’s handkerchief, somebody’s penknife, torn newspapers, coins, faded flowers picked in another meadow.”

“I see. A roadside picnic.”

“Precisely. A roadside picnic, on some road in the cosmos. And you ask if they will come back.”

I think this thought experiment is an interesting and relevant lens to understand what AI is doing to science, most prominently right now in mathematics. The labs are solving millennium prize problems and dumping discoveries on the field in troves so big they have to provide explainers on how to navigate the dump itself. All this new knowledge is arriving at a pace much greater than the field’s ability to process it and in a manner completely orthogonal to the operating norms of the field, wreaking havoc as a result.

Many are understandably vexed about the way in which the labs are developing knowledge and engaging with the scientific community. At the same time, everyone kept asking some version of “when will AI discover something useful?” Well, we’ve arrived (or AI has arrived?). These “discovery dumps” will only accelerate and come for every field where things can be quickly verified in a closed loop, which is what AI is extremely good at. Now we have to figure out how to integrate this new knowledge quickly enough into our model of the world.

I posted about this yesterday and, as one does, I got into an argument with a stranger. Not my preferred way to spend time online, but this time it actually contained a kernel of an interesting observation: answers will now increasingly arrive before the understanding. This is far from a new phenomenon, but it becomes a different beast when the rate at which discoveries arrive skyrockets. To quote Sakeeb Rahman, who also provided some color in the discussion:

The shape of things to come: proof will become cheap and understanding will become the bottleneck.

This whole situation then posits a question: will the rate of scientific discoveries get so rapid that we’ll completely lose the ability to wrap our heads around it, turning AI output into the proverbial alien artifacts? Artifacts that sow chaos in the wake of their arrival and by their mere existence (when discussing the most recent drop of AI scientific papers, the NYU math professor Tristan Buckmaster said that entire research programs were wiped out overnight). In other words, will the rate of discovering answers split off from the rate of our understanding?

I’m certainly not a science historian nor a philosopher, so I’m not well-placed to speculate about the answers or the exact implications. But it’s nonetheless an interesting lens for examining AI’s ongoing impact on science. This may also be one of the few valid anti-accelerationist arguments: if AI vastly outruns our ability to understand the significance of what it’s doing, the collateral damage may outweigh any pure scientific value.

Poignantly enough, the “Roadside Picnic” book explores human irrelevance and the resulting emotional toll behind this alien visit. You could speculate that, at best, aliens didn’t even notice the humans and, at worst, just didn’t care. The big difference is that an alien visit is entirely out of our control, unlike AI development. I’m fairly skeptical that any coordinated action is even possible in frontier AI development simply due to its game theory dynamics. But the labs have a choice in how they engage with scientific communities to make sure the understanding keeps up with the rate at which answers arrive and that their models act to augment scientists instead of speed-running them to history-making discoveries. Otherwise, we may become the insects watching in horror as everything we know becomes a site of one big alien picnic.