Unconvinced there’s a big result. “The logical induction algorithm that we provide is theoretical rather than practical.”
-
-
Replying to @Meaningness @abramdemski
I prodded some of the authors until one linked empirical stuff they've done. https://github.com/GallagherCommaJack/markets …
2 replies 0 retweets 2 likes -
Replying to @othercriteria @abramdemski
Can you summarize? (I’m not motivated to do work on this unless/until someone convinces me it’s exciting)
2 replies 0 retweets 0 likes -
Replying to @Meaningness @abramdemski
I'd like that summarization too! Can't grok how the market works and how to build traders without actually trying.
2 replies 0 retweets 0 likes -
Replying to @othercriteria @abramdemski
A thing that is like this, that actually works and is well-understood, is https://en.wikipedia.org/wiki/Probably_approximately_correct_learning …
2 replies 0 retweets 0 likes -
Replying to @Meaningness @othercriteria
PAC learning doesn't get all the nice properties this algorithm does, but yes it's similar.
1 reply 0 retweets 0 likes -
Both this and PAC learning are general frameworks for bounded rationality in a machine-learning context.
1 reply 0 retweets 1 like -
The criteria in the paper defines "good behaviour" in a wider case, ie, cases that aren't PAC-learnable.
1 reply 0 retweets 0 likes -
Replying to @abramdemski @othercriteria
Yes… but are there any cases in which it’s practical? If not, then may be interesting as logic, but not AI.
3 replies 0 retweets 0 likes -
Replying to @Meaningness
Besides... Probability theory isn't poly time. Logic isn't poly time. Why should the unifying theory be poly time?
1 reply 0 retweets 0 likes
If it’s AI, I may be interested now. If it’s logic, I’ll wait until logicians tell me it’s a result. Not holding my breath!
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.