Part 2: Have the Superbirds Arrived? Are They Taking Over?
Dr. Avian now claims that his work with trained birds show that intelligence does not require inner models or internal representations, as formerly thought(Here’s Part 1 of the story: Move over, AI. Bird brains are giving you a run for your money.)
Ignoring the staff complaints (which, he insists, are overblown), Dr. Avian focuses on the data: efficiency metrics for training are up, cost per birdhouse is down, and the newest generation of Coordinated Avian Models (CAMs) — trained on the procedural outputs of the previous generation — show signs of something new. The feeders are getting stranger.
The trained birds’ products are not just hybridized or randomized, but seemingly intentional in their irregularity. One includes a false opening. Another uses asymmetry to deflect wind. A third looks like it was built for a species of bird no one has ever seen.
Avian publishes a white paper: “Toward Emergent Avian Intentionality: Design Without Designers.”
The media runs with it. The term “superbird” appears in WIRED.
The Avian Scaling Hypothesis
On the heels of his growing commercial empire, a Congressional inquiry into potential military applications, and a wave of breathless attention from AI boosters, Dr. Avian delivers what he later calls his “master stroke.”
Image Credit: Nobilior - In a series of high-profile interviews, grant applications — and increasingly unhinged social media posts — he announces that his research has not only demonstrated scalable intelligence but proven the inevitability of artificial general intelligence (AGI). He now claims that his work with trained birds show that intelligence does not require inner models or internal representations, as formerly thought. It doesn’t require language or abstraction. It doesn’t require a theory of mind.
It requires scale. And birds. Lots of birds. Pundits previously gaga over large language models (LLMs) in AI now feverishly hype and debate the future of CAMs.
Avian goes on Joe Rogan and chortles “A sufficient number of well-coordinated bird brains, that we can assume individually know nothing of the designs, can rival human architects in the depth and beauty of their creations.
Rogan squints. “Bird brains?”
You heard him right.
In fact, as he points out, the logic of his Avian Scaling Hypothesis (ASH) just mirrors the mainstream Scaling Hypothesis in AI: more data plus more parameters plus more compute equals more intelligence. No one should be confused.
But the ASH seems to push that logic into absurdity: Just keep adding birds. Eventually, the flock learns to reason.
CAMs are getting in the way of the AI conversation. MIT’s Lex Fridman wonders aloud whether the field is now committed to “bird logic.”
In one widely circulated interview, he asks Stuart Russell if “More data, more layers, more models, more training” just amounts to the same thing.
Image Credit: Alex from the Rock - Russell, visibly annoyed, pushes back: “AGI doesn’t magically emerge — like it seems to with CAMs. It’s in the neural network.”
“CAM intelligence is in the network too,” Lex counters.
Russell sighs. “Well, humph… neural network nodes aren’t bird brains.”
“They’re not even bird brains,” Lex interjects. “Birds are smart!”
“That’s not really the point, you see. I think… I think CAMs are a fraud, actually. And they’re not like neural nets because we’re comparing out of scale. There are orders of magnitude more nodes.”
On and on it goes — circling the real question: Does coordination alone constitute understanding? Fridman keeps cutting Russell off whenever the latter tries to “intelligence-smuggle” his artificial nodes into consciousness.
“We’re attributing cognition to the system in both cases,” Lex says. “And we don’t actually know, in either case, whether it has intelligence — or just simulates it.”
He leans forward. “Step back, and you see output that looks intelligent. Isn’t that the whole point?”
Dr. Avian, in the meantime, is busy getting rich.
And the results, he argues, speak for themselves. The number of functional birdhouse designs continues to rise. Using a mix of existing plans and public databases of known models, he introduces controlled randomness into the training cycles. The system responds with architectural variations that even he finds surprising.
Some of the designs appear to serve new ecological functions. A few match nothing in the IKEA archives or any known design catalog.
One is eerily beautiful.
He begins speaking at conferences under a new banner: “Intelligence without mind.”
While Sam Altman and other AI enthusiasts remain cagey about when mind might emerge as AI gets more intelligent, Avian is perfectly clear: there is no mind at all in CAMs. And yet, they’re behind a staggering number of new designs.
It’s a master stroke, all right. A TED Talk is scheduled. Investors circle. Defense contractors request a closed-door briefing.
And all the while, the birds keep building, building…
Next: Part 3: A Wren arrives— and ruffles many a feather.
Here are all three parts of my thought experiment, told as a story:
Part 1: Move over, AI. Bird brains are giving you a run for your money. Could ten thousand birds develop a theory of mind just by scaling? A tale in three parts Dr. Avian was sure that he had found a formula for intelligence without anything like a human mind, and his program appeared to be working.
Part 2: Have the Superbirds arrived? Are they taking over? Dr. Avian now claims that his work with trained birds show that intelligence does not require inner models or internal representations, as formerly thought. Avian is perfectly clear: There is no mind at all in Coordinated Avian Models (CAMs). And yet, they’re behind a staggering number of new designs.
and
Part 3: A Wren arrives — and ruffles many a feather. Dr. Wren, a cognitive scientist, identifies a problem with assuming that adding another ten thousand pigeons to the project will produce novel designs… We remain confronted by the same old mystery: who, or what, imagines the birdhouse in the first place?
