Neurology Artificial Intelligence Program

@DavidJonesBrain of @MayoClinicNeuro is discussing imaging approaches to distinguish forms of dementia at the morning session of the 2023 Holloway Summit.

Mayo Clinic embarks on a $5 billion project to revolutionize its flagship medical campus in #RochMN. Five new buildings will fuse traditional medical services with AI and digital tools, creating a 21st-century vision of clinical care. bit.ly/3uE2Wuk @statnews

Yes, I do. LLMs produce their answers with a fixed amount of computation per token. There is no way for them to devote more (potentially unlimited) time and effort to solving difficult problems. This is very much akin to the human fast and subconscious "System 1" decision process. True reasoning and planning would allow the system to search for a solution, using a potentially unlimited unlimited time for it. This iterative inference process is more akin to the human deliberate and conscious "System 2". This is what allows humans and many animals to find new solutions to new problems in new situations. Some AI systems have planning abilities, namely those that play games or control robots. Game playing AI systems such as AlphaGo, AlphaZero, Libratus (poker), and Cicero (Diplomacy) have planning abilities. These systems are still fairly limited in their planning abilities, compared to animals and humans. To have more general planning abilities, an AI system would need to possess a *world model*, i.e. a subsystem that can predict the consequences of an action sequence: given the state of the world at time t, and an imagined action I could take, what would be the set of plausible states of the world at time t+1. With a world model, a system can plan a sequence of action so as to fulfill an objective. How to build and train such world models is still a largely unsolved problem. Even more complex is how to decompose a complex objective into a sequence of sub-objectives. This would enable hierarchical planning, something that humans and many animals can do effortlessly but is still completely out of reach of AI systems.

Tau PET is an attractive biomarker for Alzheimer’s disease but is not readily accessible to patients. Lee et al. present an AI model that can impute tau PET images from more widely available cross-modality imaging inputs. tinyurl.com/mr38rbmc

Mayo Clinic in Rochester, MN has been ranked #1 in the World's Best Specialized Hospital 2024 rankings in both Neurology and Neurosurgery! bit.ly/45dNnpT

We're looking for a google cloud engineer to work with our interdisciplinary team at @MayoClinic. We're already well on our way to deploying an ecosystem that will virtualize the Neurological exam and give providers tools to view and explore AI analyses. tinyurl.com/ypn2py97

There are two paths to "understanding reality". (1) LLM: Take everything written about reality and write a program that answers questions about it. (2) CI: Take a tiny chunk of reality that you understand well, and ask: What would it take for a machine to reach YOUR level of understanding. I call (2) "Understanding understanding", and I doubt whether (1) can succeed without (2). The main challenge of "understanding reality" is to take what we know about (2), which is quite a lot, and harness it to guide (1). @elonmusk, @xAI,@GaryMarcus,@lexfridman,

1/ Excited to share a new NeuroAI preprint: “What can 1.8 billion regressions tell us about the pressures shaping high-level visual representation in brains and machines?” Led by Colin Conwell, with @_jacobprince_, @cvnlab, and @grez72. biorxiv.org/content/10.1101/…

We are visualizing our data all wrong. Typically, academic graphs show statistical confidence intervals or standard errors. This paper shows that everyone from doctors to tenured faculty misread these graphs, assuming effects are larger. Show variability! jakehofman.com/pdfs/illusion…

Here are the areas of hypometabolism on one of my patients. What are the symptoms? #endalz #getanFDG