NeuroAI Paper Sparks Debate: Is AI-Brain Convergent Evolution Inevitable or Just an Analogy?

A recent NeuroAI paper proposes that "contravariance" might explain the convergent evolution between AI models and biological brains, arguing that such convergence is mathematically inevitable under certain biological assumptions. This claim has sparked intense academic debate over whether "convergent evolution" truly applies to AI.

Unconfirmed

Grady Booch is the primary skeptic of this theory. He emphasizes that AI models and brains operate under completely different reward functions and external pressures, with model evolution constantly driven by humans. Therefore, he views the similarity as a vague, evidence-free, and exaggerated "interesting analogy" rather than proof of genuine "convergent evolution" between biological and artificial neurons. In response to Booch's doubts, user dyamins countered by asking: if two systems exhibit parallel structures, doesn't that inherently meet the definition of convergent evolution?

Why it matters

Despite disagreements at the neuroscience level, some researchers focus on its engineering implications. Phillip Isola points out that identifying common structures across different networks holds practical value for machine learning itself, helping to design better network interfaces and architectural integrations.

2026-07-23 ~ 2026-07-23 · 7 related posts

Primary sources

NeuroAI paper argues artificial networks and real brains may converge mathematically — aran_nayebi · 2026-07-23 00:27 (UTC+08:00)
Grady Booch says AI–brain “convergent evolution” is only an amusing parallel — Grady_Booch · 2026-07-23 00:48 (UTC+08:00)
Researchers argue AI convergence patterns could still help design better ML systems — phillip_isola · 2026-07-23 09:18 (UTC+08:00)

A new theory says contravariance may explain convergence between AI models and brains — dyamins · 2026-07-23
[source] NeuroAI paper argues artificial networks and real brains may converge mathematically — aran_nayebi · 2026-07-23
[source] Grady Booch says AI–brain “convergent evolution” is only an amusing parallel — Grady_Booch · 2026-07-23
A reply challenges Grady Booch’s line between “parallel” and convergent evolution — dyamins · 2026-07-23
Grady Booch says the AI-brain analogy is vague, unsupported, and overblown — Grady_Booch · 2026-07-23
[source] Researchers argue AI convergence patterns could still help design better ML systems — phillip_isola · 2026-07-23
Grady Booch says the “convergent evolution” claim overreaches; ML researchers still see value — Grady_Booch · 2026-07-23