Comments on: The unreasonable effectiveness of data http://www.vetta.org/2009/06/the-unreasonable-effectiveness-of-data/ Tue, 27 Jul 2010 16:18:49 +0000 hourly 1 http://wordpress.org/?v=3.0 By: Shane Legg http://www.vetta.org/2009/06/the-unreasonable-effectiveness-of-data/comment-page-1/#comment-20012 Shane Legg Sun, 11 Oct 2009 12:32:24 +0000 http://www.vetta.org/?p=475#comment-20012 I think the ideas discussed in this post, while not unknown, are greatly undervalued by the community. It's also interesting from an AGI development perspective: even if you had a the right hardware and a pretty good algorithm, it might take quite a long time and a huge amount of data before your system developed things like human common sense knowledge or the ability to model human ethical decision making. I think the ideas discussed in this post, while not unknown, are greatly undervalued by the community.

It’s also interesting from an AGI development perspective: even if you had a the right hardware and a pretty good algorithm, it might take quite a long time and a huge amount of data before your system developed things like human common sense knowledge or the ability to model human ethical decision making.

]]> By: The Inevitable Effectiveness of Data « Breakingkeyboards’s Blog http://www.vetta.org/2009/06/the-unreasonable-effectiveness-of-data/comment-page-1/#comment-19862 The Inevitable Effectiveness of Data « Breakingkeyboards’s Blog Wed, 24 Jun 2009 19:23:49 +0000 http://www.vetta.org/?p=475#comment-19862 [...] Inevitable Effectiveness of Data Reading the post of Shane Legg “unreasonable-effectiveness” and reading the related article “The Unreasonable Effectiveness of Data” by Alon [...] [...] Inevitable Effectiveness of Data Reading the post of Shane Legg “unreasonable-effectiveness” and reading the related article “The Unreasonable Effectiveness of Data” by Alon [...]

]]> By: Denis http://www.vetta.org/2009/06/the-unreasonable-effectiveness-of-data/comment-page-1/#comment-19859 Denis Tue, 23 Jun 2009 11:20:26 +0000 http://www.vetta.org/?p=475#comment-19859 Oh Yes! this is what I try to explain in my blog and what I am using in my predictor project. How can human solve exponential problems? There are unsolvable problems, why can human solve these problems? Becouse all existing problems are restricted by the reality. Using logic and mathematics you can construct problems which need exponential resources , but these resource does not exist. I think is possible to use huge dataset to construct an about "real general priori knowledge" to speed up algorithms giving a general context. If we start from the point that all the possible 2^1000 bit strings does not exist knowing which of these bitstrings of length 1000 exist is not an useless information . Oh Yes! this is what I try to explain in my blog and what I am using in my predictor project.
How can human solve exponential problems? There are unsolvable problems, why can human solve these problems?
Becouse all existing problems are restricted by the reality.
Using logic and mathematics you can construct problems which need exponential resources , but these resource does not exist.
I think is possible to use huge dataset to construct an about “real general priori knowledge” to speed up algorithms giving a general context.

If we start from the point that all the possible 2^1000 bit strings does not exist knowing which of these bitstrings of length 1000 exist is not an useless information .

]]> By: Roko http://www.vetta.org/2009/06/the-unreasonable-effectiveness-of-data/comment-page-1/#comment-19845 Roko Sat, 20 Jun 2009 21:28:13 +0000 http://www.vetta.org/?p=475#comment-19845 Great post, shane. Thanks for the link to the article by Norvig, too, that will be helpful for my research. "So, follow the data. Choose a representation that can use unsupervised learning on unlabeled data, which is so much more plentiful than labeled data. Represent all the data with a nonparametric model rather than trying to summarize it with a parametric model, because with very large data sources, the data holds a lot of detail." Great post, shane. Thanks for the link to the article by Norvig, too, that will be helpful for my research.

“So, follow the data. Choose a representation
that can use unsupervised
learning on unlabeled data, which is
so much more plentiful than labeled
data. Represent all the data with a
nonparametric model rather than trying
to summarize it with a parametric
model, because with very large data
sources, the data holds a lot of detail.”

]]> By: Kevembuangga http://www.vetta.org/2009/06/the-unreasonable-effectiveness-of-data/comment-page-1/#comment-19844 Kevembuangga Sat, 20 Jun 2009 18:18:38 +0000 http://www.vetta.org/?p=475#comment-19844 Why don't you use that method to have AI learn about AI? Oh! I forgot, this is the very goal of seed AI. May be you are missing some ingredient? :-D Why don’t you use that method to have AI learn about AI?
Oh! I forgot, this is the very goal of seed AI.
May be you are missing some ingredient? :-D

]]>