Dear Tim,
Thank you for that information. I am with you on your main critique.
Do you think that Legg's proof is somehow related to the "no free lunch"
theorem in optimization theory?
The two combined seem to point quite strongly that:
- there never will be a silver bullet algorithm for learning or for
prediction
- the "solution" to strong AI will be complex
Am I stretching it if I expand the conclusions even further by stating:
- strong AI needs to be so complex that human developers cannot
understand it sufficiently to "program" it? (Legg also points out that
complex algorithms cannot be analyzed due to Goedel's incompleteness.)
If that sounds correct, would it be fair to say that then necessarily
this follows:
- So, in theory, strong AI can only be evolved ?
And if all of the above was right, is there any alternative but seeking
methods to accelerate the evolving procedures (such as e.g.,
AI-Kindergarten)?
Thank you.
Danko
On 12/02/16 04:52, TimTyler wrote:
I read and reviewed Pedro Domingos's book "The Master Algorithm".
My review is here:
http://smile.amazon.com/gp/cdp/member-reviews/AYJ8P83FHQARZ
To summarize my biggest criticism:
The book documents the search for a silver bullet of machine intelligence.
The author didn't seem to be familiar with Legg's 2008 proof that machine
intelligence will be complex. In "Is there an Elegant Universal Theory
of Prediction?" Legg offers a simple, constructive proof that, for any
prediction algorithm, there exist sequences with similar Kolmogorov
complexity to the prediction algorithm, that the predictor can never
learn how to predict. Legg's conclusion is that successful general
purpose predictors of complex sequences will themselves necessarily
be highly complex. Machine intelligence doesn't have a silver bullet.
--
__________
|im Tylerhttp://timtyler.org/
*AGI* | Archives <https://www.listbox.com/member/archive/303/=now>
<https://www.listbox.com/member/archive/rss/303/27953441-c5c84d1c> |
Modify
<https://www.listbox.com/member/?&;>
Your Subscription [Powered by Listbox] <http://www.listbox.com>
-------------------------------------------
AGI
Archives: https://www.listbox.com/member/archive/303/=now
RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-f452e424
Modify Your Subscription:
https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-58d57657
Powered by Listbox: http://www.listbox.com
