My former teacher Robin Hanson argues that the most likely way forward for artificial intelligence is whole brain emulation, the process of copying a brain neuron-for-neuron into a computer simulation. Here are Robin’s posts on “ems“; here are my previous posts, which present objections to Robin’s hypothesis. In this post I’ll present one more.
I have become increasingly aware lately of how maladapted humans are to our current environment. We are apes from the Paleolithic Era. Yet here we are, eating grains and candy, sitting indoors at desks all day.
Robin has discussed our maladaptations before, but he seems to focus on cultural values and norms. In Robin’s terminology, there are forager values and farmer values. In the industrial age, as we grow wealthier and social pressure subsides, we retreat to more forager values. Robin speculates that a post-industrial em civilization may have more use for farmer values.
Culture evolves rather rapidly. Physiology evolves much more slowly, and for ems (which may be infinitely-lived) it evolves not at all. A big concern for ems, therefore, is health. Since ems don’t have physical bodies, we don’t need to worry much about cancer and heart disease. Rather, we need to address the problem of mental illness.
Is our industrial environment suited for the flourishing of our predominantly Paleolithic minds? I think the answer is that it’s clearly not optimal. Fifteen percent of people in developed countries are severely depressed; that includes four percent of preschoolers, who constitute the fastest-growing market for antidepressants. One of my new favorite blogs is Evolutionary Psychiatry, which documents the connection between modern diets and practices and numerous mental illnesses such as schizophrenia and dementia.
To some extent, we can use this information to help ems. To whatever extent that we need to simulate a diet for them (they won’t need food for energy), we can base it off of Paleolithic foods. Maybe we can be sure that their brains receive whatever signals our brains receive when we get copious amounts of sunlight and exercise.
But even this misses the point. However large a change it is to go from a hunter-gatherer environment to a modern industrial environment, it is a much larger change for a mind to go from the current environment to a disembodied existence. Ems will almost certainly go insane, no matter how much we try to help them. For every environmental factor for which we control, there will be perhaps thousands we have never considered.
This is of course not an argument against the technical feasibility of copying a mind into a machine. We may succeed at doing so. But if I am right, this accomplishment won’t change the world. Instead of getting cheap labor, we will get the experience of watching a human mind go rapidly insane. I am skeptical that subsequent iterations, if they are even allowed given the ethical considerations, will fare much better. It seems, therefore, that the singularity will have to wait for weak AI, which Robin says is much more difficult, to develop.