Campus News
The frontier of brain science: AI and organoid research takes center stage in Silicon Valley
At a meeting of minds, researchers discussed a platform that could finally reveal the intimate workings of the human brain, and what happens if we succeed.
The technology to truly understand the human brain is almost here. Within two years, researchers could have working prototypes of a platform that combines artificial intelligence with living brain tissue to decode how neural circuits form, function, and fail. Building this technology is only the beginning, however. Turning it into a tool that transforms medicine will require an unprecedented collaboration across multiple sectors, and serious engagement with the ethical questions it raises.
In mid May, an intimate gathering at the Rosewood Sand Hill in Menlo Park hosted by George and Rafe Kraw brought together some of the people who could make that collaboration happen.
“Using AI and human brain organoids together to understand how our brains actually work will give us a much deeper understanding of ourselves,” said David Haussler, Scientific Director of the UC Santa Cruz Genomics Institute. “And I want us to really engage with the question of what happens if we succeed.”
Organoids that learn, paired with powerful AI
Brain organoids are tiny, three-dimensional structures grown from human stem cells that mimic aspects of the developing brain. After a decade of work and tens of millions of dollars in philanthropic and government investment, the Braingeneers, an interdisciplinary research group at the UC Santa Cruz Genomics Institute, UCSF, UCSB, Stanford, and Washington University St. Louis, has developed technology to grow these structures at scale, record their electrical activity, and train them to perform simple tasks.
At the event, attendees watched a human brain organoid that had been hooked up to a virtual environment in the Braingeners lab for about half an hour and was learning to play a game in which it balanced a pole on a cart. Ash Robbins, a postdoctoral scholar with the Braingeneers who was running the demonstration, handed a keyboard to a volunteer in the audience to have them try the same game. The volunteer struggled to match the organoid’s performance, drawing a few laughs.
“This is just to demonstrate that it is actually a very hard task,” Robbins said. “There is no way the organoids are doing it by accident.”
In other words, the researchers had successfully been able to use reinforcement learning, similar to that used in AI, to get a lab grown human brain tissue to solve a tricky and well-known task in mechanical control theory.
A second demonstration led by Yohei Rosen, an MD and postdoctoral scholar with the Braingeneers, introduced the audience to both standard and vascularized organoids. Participants watched an “Organoid ER” livestreamed from the Braingeneers lab in Santa Cruz, in which a lab member administered a drug to induce a seizure. The audience then observed a hierarchy of AI models, which the team had playfully labeled with “Intern,” “Resident,” and “Attending Physician” personas from the hit TV show The Pitt, as these AIs recognized the seizure in the brain organoid and successfully administered a drug to calm it.
From Prototype to Platform
The Braingeneers shared their ambition to scale up, building a system to grow thousands of organoids simultaneously and use AI to learn from them how neural connections develop and where they can go awry. The group’s recent selection as the leaders funded to build the AI-driven Brain Cell Data Explorer for the NIH-wide BRAIN Initiative positions them as a hub for this emerging field, building on their ongoing role as the Analysis Center for the Psygene project from the National Institute of Mental Health. The latter project will investigate the function of the 250 human genes most strongly linked to neurodevelopmental and neuropsychiatric disease, using both human organoids and experimental animals.
The Braingeneers’ full vision, a platform capable of conducting hundreds and soon thousands of organoid experiments in parallel over months, will require partners: funders, technology collaborators, and experts willing to help shape how this work unfolds.
Central to this effort is a collaboration with Jure Leskovec, Professor of Computer Science at Stanford University and one of the most influential figures in modern AI. Leskovec pioneered graph neural networks, a technology now used by Facebook, Pinterest, and Amazon, and he is a world leader in AI for the Life Sciences. At the event, he presented his lab’s work on foundation models for biology, AI systems that can learn universal AI representations of biological cells and tissues across tissues and species. He explained that his lab is training systems that will become as fluent in the language of cells as LLMs like Claude and ChatGPT are in human dialogue.
Combined with the Braingeneers’ organoid platform, these tools could help researchers understand how neural circuits develop, respond to drugs, and go wrong in disease, all at a speed and scale previously impossible.
A conversation about how to proceed
The audience members were not casual observers. These were people who had dedicated their careers to pieces of this puzzle. Bob Klein, who spearheaded California’s $3 billion investment in stem cell research through CIRM, was among the attendees. So was Sara Shnider, Senior Director of the One Mind Accelerator, an organization dedicated to accelerating treatments for brain diseases; Stephanie Morris, Director of Life Sciences at the John Templeton Foundation; and Evan Kotler, Director of Research and Health Security at the Helena Foundation. Prominent researchers from a number of different universities and institutions debated what comprehension of human neural circuitry could mean for the neurological and psychiatric conditions that affect billions worldwide.
But the evening was not just about the science. Haussler asked attendees to rearrange their chairs into a circle and asked people to discuss their feelings about the potential ramifications of this work for both health and humanity. He started off the discussion by having Hank Greely, Director of Stanford’s Center for Law and the Biosciences and a leading authority on the ethics of emerging biotechnology, discuss what guardrails this technology should have.
Under Chatham House rules, attendees shared their feelings freely. Some raised concerns about governance, unintended consequences, and the concentration of power as AI becomes more powerful, especially as it interacts directly with human brain tissue. Others continued the discussion on health research, noting that organoid research has already contributed to a potential treatment for epilepsy and other disorders, and emphasizing that a deeper understanding of neural circuits could transform care for the billions affected by neurological and psychiatric conditions. It was widely acknowledged that ethical guardrails must be developed alongside the science for experiments where advanced human brain organoids interact with advanced AI systems. Continued partnerships with ethicists, philosophers and social scientists in the responsible development of this technology was advised.
What comes next
The gathering marked the beginning of a broader effort to build the larger partnerships this project requires. UC Santa Cruz is positioning itself not just as a research leader, but as an organizing force, bringing together the collaborators, funders, and thinkers needed to realize a vision too large for any single institution.
“This is the beginning of something very big,” Haussler said. “It’s too big for any one institution to take on the whole project, but we’re working to coordinate partnerships that can reshape the future in a way that benefits us all.”
To learn more about the Braingeneers, visit braingeneers.ucsc.edu or contact genomics.info@ucsc.edu.