Introducing AI and Therapy
From BVS.net
Your friends at Boston Venture Studio are now “venturing” to create products at the intersection of AI chat bots and psychotherapy.
This is a rapidly evolving field, with new studies coming out every week!
Today, already there are millions of people who are using Chat GPT out of the box as their therapist. Some studies have shown over 50% of people in their 20s trusting AI chat more than a human therapist.
Also, there is more demand for human therapists than there are therapists, and many people do not have health insurance which covers psychotherapy. And in many cases, paying privately can cost $200 an hour or more — a price that many people can not afford to spend each week.
Some people are using AI therapy in conjunction with a human therapist, using the AI “between sessions”.
While AI therapy is helping millions of people each week, there have been some horrible incidents, where some AIs have hallucinated and talked with people about how to kill themselves. At the same time, AI therapy has saved the lives of many other people.
For AI therapy to really take off, we clearly need much stronger guardrails - honoring the First Law from Isaac Asimov’s “Runaround” from 1942—
“A robot may not injure a human being or, through inaction, allow a human being to come to harm.”
We at BVS are working to create those guardrails in the AI therapy products we produce.
We might not also call our AI products “therapists” so as to not confuse a patient that an AI can truly do all that a trained human therapist is allowed to do. Stay tuned for more on that.
Meanwhile, welcome to “AI Therapy” substack by BVS.net
Below you will see sections from a Linkedin Article(Martin Bello)
3. Mental Health Diagnostics Using fMRI Data and AI Models Using fMRI datasets, I designed a multi-stage pipeline for preprocessing, feature extraction, and classification of neuropsychiatric conditions such as ADHD, bipolar disorder, and schizophrenia. This includes feature selection using Joint Mutual Information, state-space modeling, and ensemble MLP classifiers—all developed with direct assistance from ChatGPT. The models show promising discriminatory power and could enhance early-stage diagnostics.