HN Explorer

Ask HN: Pure mathematical route to intelligence?

23 points by friendoamano ↗ HN
Are there purely mathematical inspired routes to Artificial intelligence. Such routes perhaps assume intelligence as a property of some exotic pure mathematical structure? Such route deviate from the human-brain/nature inspired ways of intelligence

17 comments

[ 1.6 ms ] story [ 50.6 ms ] thread
mindcrime
The closest thing I can think of would be Algorithmic Information Theory[1]. Marcus Hutter has done some work with that vis-a-vis Artificial Intelligence[2][3][4]. And while it's not quite the same thing, you might find Pei Wang's NARS[5] approach of interest.

[1]: https://en.wikipedia.org/wiki/Algorithmic_information_theory

[2]: http://www.hutter1.net/ait.htm

[3]: http://www.hutter1.net/ai/uaibook.htm

[4]: https://en.wikipedia.org/wiki/AIXI

[5]: https://cis.temple.edu/~pwang/papers.html

friendoamano
Thanks for the links especially the Marcus Hutter links
shahbaby
I find questions like this frustrating...do you know what intelligence actually is? I don't mean just a vague notion of it but an understanding of the principles behind which it works.

Nature had billions of years and a possibly infinite universe to figure it out. You don't have that long.

friendoamano
I think if nothing else, as long as I can see it do the things(on it's own) what a human being can do to the point where human intellect and creativity is completely devalued, that would satisfy my definition of intelligence
mindcrime
But nature wasn't trying to create intelligence. Intelligence emerged because intelligent creatures lived long enough to pro-create and spread their genes.

Yes, evolution is a powerful force, but there's no particular reason that I'm aware of to think that we, as human, without our own intelligence and reasoning abilities, can't engineer other intelligences.

ADanFromCanada
"But nature wasn't trying to create intelligence"

That may seem true... but what if it was/is though?

muzani
That's the whole argument around theism.
openfuture
Paraphrasing a quote from Al Perlis:

“there is no formal way from informal to formal”

aserafini
The question reminds me of a paper I read: Causal Entropic Forces, sometimes referred to as 'entropic intelligence' but the basis is physics rather mathematics.

https://physics.aps.org/articles/v6/46

In this theory, intelligence is equivalent to maximising future possible states in a system. There are some interesting videos on YouTube about it as well.

kleer001
IMHO Game Theory is the key to that
muzani
Game theory will result in something weird like Facebook, which is marginally useful, but comes with so many downsides that people wish they didn't use it.
kleer001
I'm interested in how you got to that conclusion.
alacer
Check out the work of Dr. Sergio Pissanetzky on artificial general intelligence starting at http://sergio.pissanetzky.com/ and http://sergio.pissanetzky.com/index-2.html. It seems plausible and promising, but details are omitted preventing easy verification and it would have been convenient if he used graphs insted of causal sets. Additional references to his work are in his Google Scholar page at https://scholar.google.com/citations?user=KEuEW8QAAAAJ&hl=en, in particular see his 2 most recent publications near the bottom.
hsikka
This is interesting. I’ve opted completely for the other side of the spectrum, modeling networks rigorously to the biological standard, but this may be worth looking into.
whatsstolat
The universe seems to run on a maths based OS so I'd think so.
kleer001
Intelligence seems to me a particularly adequate solution to the existential problems of being and as such as inevitable as life its self, being a local entropic minima. To put it another way I think you can extrapolate life from physical laws and intelligence from life.