


Scientific Director and President at Laureate Institute for Brain Research and Professor of Psychiatry at UCSD


Beautiful work by Haixia Zheng ( laureateinstitute.org/haixia… ) showing that CMV-associated T-cell differentiation may be a candidate pathway to depression risk: doi.org/10.1016/j.bbi.2026.1…

This was a great team effort with @royperlis @CRodriguezMDPhD @JohnTorousMD @KhalsaLab @pshrink and karthik.sarma@ucsf.edu based on the workgroup meeting at #ACNP in January of 2026, there are some useful directions and suggestions here: rdcu.be/fA4UR

To mitigate automation bias, implement a SEIG loop: 1️⃣ Specify task bounds 2️⃣ Elicit structured data 3️⃣ Interrogate assumptions 4️⃣ Ground claims in the record LLMs function as bounded hypothesis generators, not autonomous actors. #ComputationalPsychiatry #ClinicalAI #HealthTech

In a new viewpoint piece, jamanetwork.com/journals/jam… deployment of LLMs in psychiatry requires strict oversight to @JohnTorousMD and I argue that mental health practitioners need specific training to prevent fluent but fabricated clinical narratives.

How do anxiety disorders, OCD, and PTSD change our behavior? Our neuroimaging research, out now in Molecular Psychiatry, shows: it's about more than the diagnosis. ⬇️ Learn more: doi.org/10.1038/s41380-026-0…

I'm excited to share our new paper published today in JAMA Psychiatry: jamanetwork.com/journals/jam…



AI is becoming clinically relevant for mood and anxiety care—but the evidence is uneven, and enthusiasm must be matched by guardrails. I point out that the priority now is not “AI everywhere,” but actionable, validated, clinically accountable tools that improve care.

We have less than 1 week until the Embodied Minds Summit, and the agenda is finalized! embodiedminds.ai/ If you are still waiting for an invitation, the final wave for this small invite-only meeting will be going out later today, so stay tuned.

The draft agenda for the UCLA Embodied Minds Summit is now live on the website! See “Day 2 — The Conference”. It contains the workshop assignments and speaker order. embodiedminds.ai/

At the Annual UNL Symposium on Motivation



This is what I see
GPT-5 Pro just scored a 148 IQ on the official Mensa Norway test that’s genius territory. This isn’t a cherry-picked result either… it’s the average over multiple runs. We’re now looking at an AI that outperforms almost every human on Earth in raw problem-solving ability. The “AI will never match human intelligence” crowd just got left in the dust.