Discussion about this post

User's avatar
Daniel Van Zant's avatar

One aspect of this connectomics progress that my PhD dissertation is centered around is what connectomics + large language models means for the future of brain research. The richness of connectomics data allows you to perform purely computational hypothesis testing. This means you can come up with a theory, and come up with a structurally-based testable hypothesis, and then test it, without ever having to put in the time and resources for a physical experiment.

The current crop of large language models is rapidly advancing in their ability to generate bioinformatics code, and useful hypotheses. This means that in the future you will be able to propose some high-level theory about how some part of the brain functions. Then you would press a button and instantly have 100 or 1,000 hypotheses tested, get a report of which hypotheses have falsified your theory, and then go back to the drawing board and build a more robust theory, all without expending significant time or resources.

This would be huge for the acceleration of brain research. I am building out a proof-of-concept of this system for the fruit fly connectome, but my long-term career-level goal is to be able to implement this system on the mouse, and eventually human connectome as the data continues to be developed.

Expand full comment
Ulkar Aghayeva's avatar

great piece! one question:

> The resultant images — black and white and just a single electron in width — prove challenging to read.

can this be true? an image surely can't be a single electron in width

Expand full comment
4 more comments...

No posts