Researchers pitch AI to accelerate Alzheimer's discovery while urging guardrails and validation

Wendy Nielsen of the National Science Foundation told the council that AI can bring broad literature, multimodal data, and hypothesis generation to Alzheimer's research, and described a major open-science investment (an AI-for-science initiative) that provides compute and toolkits for academic researchers. "AI isn't gonna cure everything," she said, but it can aggregate literature, support hypothesis generation, and scale analyses that humans cannot. Nielsen emphasized open-source platforms and the need for scientist-AI partnerships and human oversight.

Dr. Randy Bateman described c-Brain, a pre-competitive consortium that aggregates longitudinal clinical data, PET scans, blood and CSF samples, and postmortem molecular mapping, then applies agentic AI tools (an "open scientist," an "insight engine," and a data analysis agent) for rapid analysis and hypothesis testing. He reported that in internal tests the system reproduced most findings of manual analyses and produced novel leads, acknowledging mistakes and stressing the need for validation: "It found 50 of our 52 findings... it made 1 mistake, and the other it found something we did not find," he said, adding that these tools must be validated and fit-for-purpose before clinical or regulatory application. Council members asked about clinical governance and whether AI should be a designated provider type for reimbursement; panelists said governance and standards are necessary but that the tools have potential to expand access to specialized assessments.