Hacker News Reader

Also predictions in biology take months or years to validate, so they lack the fast feedback loop of the vision and NLP world where the feedback is almost instant.

Combine this with the fact that In vivo data in biology is extremely limited, and we see copying the NLP and vision playbook into biology is challenging

View on HN

This. Many of the predictions we're talking about are potentially years in the making, involve expensive data collection to validate, suffer from a lot of stochastic noise, etc.

Honestly even if a prediction comes an experiment, and they know exactly how the experiment was done, it takes month to years to follow up and verify.

Generative AI is basically going to flood the field with more predictions, but with little explanation of how, and doing nothing to alleviate the downstream verification process.

And when it's off in its prediction, without an explanation of how, you have no chance to revise your prediction, it's just all the way back to square one.

Reply on news.ycombinator.com