[oldgun]

This could be the right step towards algorithm transparency. It is critical in the information age we're in, while Facebook is promoting alt-right conspiracies, and YouTube is feeding flat-earthers.

Yes. There are more technical issues to resolve. (e.g., How do we verify the algorithm they present is the actual algorithm they use?) Perhaps third party audit, APIs for third parties to test the algorithm? We need constructive criticism than cynicism.

[saagarjha]

I see another problem, and that is the one where they dump a model on the internet somewhere and whenever something goes wrong shrug their shoulders and say "well, that is what the model gave us…we don't know why; you can see for yourself how we didn't bias anything".