Hacker News Reader

Comment by Loq | Hacker News Reader

If you want to understand Kant in the language of (self-supervised) machine learning, I can highly recommend this rather astonishing PhD thesis:

https://www.doc.ic.ac.uk/~re14/Evans-R-2020-PhD-Thesis.pdf

View on HN

> "We note in particular that in the sequence induction IQ tasks, our system achieves human-level performance. This is notable because the Apperception Engine was not designed to solve these IQ tasks; it is not a bespoke hand-engineered solution to this particular domain. – Rather, it is a general purpose system that attempts to make sense of any sensory sequence, that just happens to be able to solve these IQ tasks “out of the box”.

Worth repeating:

> just happens to be able to solve these IQ tasks “out of the box”.

Will read further...

Honest question; what could anyone get out of this? What is the point of interpreting kant if the interpreter of kant can't even perceive what humans perceive? Kant doesn't seem to have much value outside of the context of human perception.

I'd probably read the intro e.g.

> 1.1.1 AI has something to learn from Kant

etc.

We share a universe of perception with that thing. If it generates the same output as humans with the same input we learn that we have learned nothing and this is also knowledge of reality.

Just started reading this thesis. I am absolutely blown away by the clarity and beauty of its writing. This is rare.

Well, it was authored in TeX, so:

https://xkcd.com/1301/

applies

This is the most staggering kind of serendipity to discover that such a thing exists. Thank you :)

Fascinating. Thank you.

Reply on news.ycombinator.com