Verified by 
In September 2023, Ilya Sutskever of OpenAI wrote the following:
"In the future, once the robustness of our models will exceed some threshold, we will have wildly effective and dirt cheap AI therapy. Will lead to a radical improvement in people’s experience of life. One of the applications I’m most eagerly awaiting."
I think we should have a measured skepticism about promises about the benefits of artificial intelligence (AI), especially from those with a professional stake in its advancement. Nonetheless, it is worth thinking through the implications of what Sutskever is anticipating. What would a world with "wildly effective and dirt cheap" therapy look like, and in what ways would it differ from our own?
Begin with our world, specifically the United States, circa the end of 2023. Here, therapy is a relatively scarce good. Telehealth promises to alleviate some of these limitations, but thus far, significant limitations remain. Accordingly, like most scarce goods, therapy is expensive (especially for consumers whose health insurance does not cover it).
For many people today, therapy is difficult to obtain and expensive. If therapy were more widely available, presumably, more people would avail themselves of it. People who now cannot afford therapy might be able to afford one hour a week. People who now limit themselves to one hour a week might opt for two hours. People needing at least two hours of therapy a week might decide they need at least four. But this process has to end somewhere. Doesn't it?
In some sense, it ends at the number of hours a week: 168. But it is worth asking whether it has any stopping place before that. The question generally does not arise in the world as we know it, the question generally does not arise, as people often cannot afford even the therapy they need.
But Sutskever's question asks us to imagine a different world where therapy is as cheap as proverbial dirt. To put the question in explicitly economic terms, what would a world look like in which the marginal cost of an hour of therapy—that is, the cost of producing one more hour of therapy, given all the infrastructure we have already built—approaches zero?
I want to suggest that the answer to this question is non-obvious. We are so accustomed to encouraging people to go to therapy, of trying to make therapy more accessible, that many of us have not reflected on the possibilities of therapy if if is abundant. Perhaps in such a world, therapy will be like water (in a developed country like the United States): Something that people use effectively and freely throughout the day without much reflection on the cost.
Or perhaps in such a world, therapy would resemble video gaming: Ignored by many, indulged in by some, and then used compulsively and excessively by a few. Would therapy be more like water or video games, or will it follow some third and perhaps novel model? This is the kind of question that reflection on abundant therapy leads us to take more seriously than we do now.
Behind this question is a more fundamental one: Would the introduction of "wildly effective and dirt cheap therapy" be, on balance, a good thing? The model of water suggests that therapy is a fundamental good that technological advances will allow us to supply ever more cheaply and effectively.
The model of video games suggests a more ambiguous lesson, as well as a cautionary tale. Of course, therapy is neither a necessity of life nor pure entertainment, but something between these suggests that unlimited therapy will be somewhat more complex in its deliverances.
As I noted at the outset, this entire discussion presupposes a certain amount of optimism about technology, particularly artificial intelligence, that merits skepticism. Some think that we are nowhere near AI being able to adequately replicate the work of a good therapist. Still, others think that we can, in principle, never even get there.
Whatever one's degree of confidence in this technological question, the thought experiment provided by AI forces a question worth asking. What would a world with unlimited therapy look like—in what ways would it be better, and in what ways would it be worse?
How much therapy would it be rational to want in such a world—an hour per week, four hours per week, or more? How much therapy, in the end, does a person need?
