If AI Thinks Like a Human It May Get Depressed
A neuroscientist considers why a serotonin function in artificial intelligence could be both a help and a hindrance.
As artificial intelligence systems become more sophisticated, we’ve grown accustomed to hearing about how they take over human domains. AI can now beat humans at Go, write clickbait and screenplays, pick out targets in drone image data, and drive our cars.
Most of these AI systems are neural networks, a type of computing architecture loosely modeled after the human brain. For now, each of these neural nets are compartmentalized and optimized for one particular task—the AI that can beat us at Go doesn’t know how to drive a car. The hope is that one day, however, these “narrow” AI systems will coalesce into an artificial general intelligence (AGI), or a computer that is able to a wide variety of tasks as well or better than a human.
Yet for neuroscientist Zachary Mainen, this raises an interesting question: If a computer is eventually able to think like a human, will it have the same sort of mental health problems?
Mainen is a researcher at the Champalimaud Center for the Unknown in Portugal and spends a lot of time thinking about depression. Depression is a relatively common phenomenon that afflicts over 300 million people worldwide and is closely linked with the neurotransmitter serotonin.
As Mainen noted during a presentation last month at the Canonical Computations in Brains and Machines conference in New York, serotonin is a neuromodulator, a specific kind of neurotransmitter used to send messages across large areas of the brain.
According to Mainen, research on serotonin has shown that it plays a large role in the brain’s ability to adapt to unfamiliar situations.
“People think of serotonin as related to happiness, but serotonin neurons appear to send a message that is not good or bad, but more ‘oops’ or surprise,” Mainen told Science after his presentation. “It seems to be especially important in breaking or suppressing outdated beliefs.”
In this sense, Mainen said depression may be seen as the brain’s inability to adapt to change. He offered an example of someone who becomes depressed after sustaining a severe injury and is unable to cope with their new disabilities. According to Mainen, drugs like selective serotonin reuptake inhibitors (SSRIs) or even psychedelics like psilocybin are able to relieve depression symptoms by fostering brain plasticity.
Artificial intelligences may need to have some sort of control mechanism built into them that functions similar to the way serotonin works in the human brain, Mainen said. This mechanism would allow the machines to adapt to new situations quickly, but it could also result in the entrenchment of certain thought patterns and result in depression in the machines as well.
“Computational psychiatry assumes we can learn about a patient who’s depressed or hallucinating from studying AI algorithms like reinforcement learning,” Mainen told Science. “If you reverse the arrow, why wouldn’t an AI be subject to the sort of things that go wrong with patients?”