[mnau]

> Here's 50 gigabytes of neural net weights.

.. from few months ago. Weights change daily, most likely updated by another NN.

I guess it's nice that lawmakers understand that at some point these companies used algorithms to search or sort stuff, but industry has already moved to another level. We might be able to explain specific result of neural networks (Shapely values or something like that), but the actual algorithm (=NN)... no way.

[Vespasian]

Well then they can explain things like what inputs to they are using (with concrete examples four each user), what metrics they are optimizing their NNs for, how their product success is measured, what their internal research is focused on etc.

I feel a lot of people on HN are looking at this from a technical standpoint while lawmakers are more interested in how these companies plan and position themselves. Explain how they "maximize profits and shareholder value" would be more accurate in my opinion