On Sat, Oct 26, 2013 at 9:31 PM, Jim Bromer <[email protected]> wrote: > Matt Mahoney <[email protected]> wrote: >> What do you figure it will take to solve NLP, in terms of computing >> power, training data, and coding effort? Or are there some fundamental >> research questions you have to answer first? > > I have no idea. I believe that using controllable experimental > platforms that can demonstrate advances is the best way to continue > advancing even if the advances are incremental. However, there are > times when you have to jump to try a different theoretical model. > > I thought you were going to ask me something about P=NP when I first > started reading your comment.
No. I guess you stopped working on the P=NP problem. I think that's a good idea because there are more productive approaches. I would recommend looking at some of the historical approaches to the NLP problem. There were hundreds of experimental programs written in the 1960's before anyone realized how hard the problem is. This will save you from having to repeat their work. Google is having some success in this field, but it isn't human level yet. For example, you can ask questions like "how many tablespoons in a cubic parsec?" and get an answer. They are able to do this because they have 46,000 employees including some top experts like Kurzweil doing AI research, and a market cap of $339 billion. Their software runs on a billion computers and phones, giving them indirect access to exabytes of personal data and exaflops of computing power. This is in addition to the million or so CPUs that they control directly, including a quantum computer. Well, I'm not sure this is helpful. Perhaps there are some fundamental questions left to be answered, but it would require some research to figure out which ones. -- -- Matt Mahoney, [email protected] ------------------------------------------- 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
