Thoughts?

Well, if your model isn't shit, you can just feed this new training data in and it should behave as before.

This whole thing only applies to modelling human behaviour on a large scale without constantly integrating new data. A.k.a. bullshit one-time models that any sane person stopped using long ago because they became increasingly unreliable over time.

Weather models won't change, they'll just input different variables regarding air pollution. Image recognition models, video tracking models, mechanical simulation models, self-driving car models, sports betting models, ... None of this will change. At worst his shitty "Top data science" company (btw that name just screams of marketing gimmick) will realise statistics and conditional sentences alone don't make a good ml.

Oh wow people during a crisis situation act differently than during a normal situation? Unbelievable.

I love it.

In 70 years: nothing going on, amazon be like: "you should buy these gas masks!"

Next day you see news about some new virus.

The future dude.

That guy sounds like he took the first two weeks of a free online AI/ML course and just spouted off as much jargon as he could remember.