Scientists Say AI Is Missing a Key Ingredient: Brain Chemicals
AI forgets what it learns. Brain chemicals might be the fix.
A group of researchers has published a paper arguing that artificial intelligence is missing a fundamental ingredient that real brains have: neurotransmitters. These are the chemical messengers your brain uses to signal things like "pay attention now," "remember this," or "calm down." In your head, learning isn't just about which neurons fire — it's also shaped by this chemical soup washing over them. Today's AI has the firing part but none of the chemistry.
Why does that matter to you? Because AI systems are famously bad at learning new things without breaking old ones. Teach a chatbot a new skill and it may quietly get worse at something it already knew. This is called "catastrophic forgetting," and it's a real headache for companies that want AI to keep improving over years rather than being retrained from scratch. Brains don't have this problem. You can learn to drive without forgetting how to ride a bike. The researchers suspect that neurotransmitters are a big reason why.
The team calls neuromodulation a "third axis" of learning — meaning it works alongside the two things AI already copies from biology: neurons firing, and the connections between them getting stronger or weaker. They argue this third piece has been almost entirely ignored in AI design, and that adding it could lead to systems that adapt and keep learning over time the way living creatures do.
Here's the honest catch: this is a "position paper." That's academic speak for an argument, not a demonstration. There's no working prototype here, no benchmark scores, no product. It's a nudge to the research community saying "hey, look over here." So don't expect your apps to get smarter because of this next week. But if the idea takes hold, it could shape how AI is built over the next several years — and that's how most big shifts in this field actually start.
- The paper argues AI borrows two tricks from brains but ignores a third: chemical signals that control when and how learning happens.
- Today's AI forgets old skills when it learns new ones — a problem called 'catastrophic forgetting' that this idea might help solve.
- It's an opinion paper with no working prototype, so the payoff is years away, not months.
Why It Matters
Could lead to AI that keeps improving without forgetting — meaning tools that get better instead of breaking.