> Up until about GPT 2, EURISKO was arguably the most interesting achievement in AI.

I agree.

> I suspect its achievements were slightly overblown and heavily guided by a human hand

So do I. We'll find out how much of its performance was real, and how much bullshit.

> the unreasonably effectiveness of differentiable programming and backpropagation has sucked up much of the oxygen in the room

The Bitter Lesson -- http://www.incompleteideas.net/IncIdeas/BitterLesson.html