MITxMGB AI Cures 2024

OREL Research Coordinator Jesse Parent recaps #AICures2024

“MIT-MGB AI Cures Conference showcases ongoing collaborative projects in clinical AI, highlighting cutting-edge research and their impact on patient care. Equally important is the process for collecting the data required to train clinical AI algorithms, with a focus on privacy, diversity, and quality.”

One of the more fascinating things about being in Boston is seeing from the peaks of certain vantage points. CBMM-10 at MIT’s Center for Brain, Minds, and Machines intriguingly held together a somewhat lopsided tripartite nexus of neuroscience, cognitive science, and computer science; yet spite of the harmonizing efforts, there was clear clout disparity in terms of what the Big AI panel would advise — particularly regarding future directions and what academia (or smaller labs in general) can offer compared to The Industry. In the case of AI Cures, there was a sense of covering the bases as well, highlighting Clinical AI, patient data applications, and Responsible AI. The message, at one point with a sort of meta-irony, was that it’s about ROI “and” saving lives — and you should be thinking about GenAI applications and enhancements. Most posters were around specific applications in these avenues, with (many students) discussing their findings, to varying degrees of novelty.

While there were many potential sound clips — albeit standard fare for a showcase event — a few quotes have stayed with me:

• “It’s a matter of priorities, and not simply theory” — Isaac Kohane, Director of the Children’s Hospital Informatics Program, Associate Professor of Pediatrics at Harvard Medical School, and an Attending Physician in Endocrinology at Children’s Hospital, Boston
• “[It’s not just a focus on longevity; ] I care how long people have agency” — Ross Zafonte, President of Spaulding Rehabilitation Network
  Earle P. and Ida S. Charlton Professor and Chair of the Harvard Medical School Department of Physical Medicine and Rehabilitation (PM&R)
  Chief of the Department of PM&R Massachusetts General Hospital
• “I did my phd in the ’80s, before the AI Winter, and I know what disappointment is — this is not what disappointment looks like” — (unknown)
• “Because of the rapid progress of AI, we’re having to reply on preprints ” (unknown)

My takeaways from the event

1. There will be no slowing down HealthAI. The final quote (intentionally unattributed) is the state of the art now, and the acceleration continues. What will slow it down, and how untethered to the past are we in terms of making products or doing research in this current regime? FDA, HIPPA, and other clearances are not going away — but where are the malleable spaces? Which leads to…
2. First-Mover Bias. It’s hard to say where GenAI and LLM bubbles will go, and when/how they will burst. But if you can remix some key trouble spots with a fairly accessible application of “AI”, there’s a major incentive to shoot your shot now. The dust may never settle, so it’s more about forging and connecting paths ahead.
3. Priorities, Theory, and Choices. Theory matters, but the burden of choices (and liability) are looming. I’m particularly curious how this will be addressed in my Data Fairness & Ethics class this term… I’m beginning to see a new frontier to the advising, consulting, or direction-setting path that our massively Big Data world has ahead. (Hello, WeRobot.)
4. Synthetic Data & Generative Visualizations. How can we create synthetic variations of otherwise hard-to-collect data? And, what are the actual use cases of GenAI imagery to help visualize or interpret such developments?
5. Sensors and Automated Data. It’s always good to see application spaces, including some of the “in home robotics” mentioned, and NETMIT Research Group & MIT Center for Wireless Networks and Mobile Computing.

Did you go to AICures2024? Are there any other Boston-area events on your radar? Later on this month there will be # MITForge2024 as part of the Imagination in Action series — I hope to see you there!

Originally published at jesseparent.wordpress.com