RE: What are useful for a phd/JOIN Dan Burfoot

From: Daniel Burfoot (d_burfoot@hotmail.com)
Date: Thu Dec 07 2006 - 02:37:13 MST


Hello all, I have been lurking on this list for a while. I wanted to comment on this thread so I figured I would introduce myself in the same message. About me: I am also a beginning PhD student, at University of Tokyo. I am primarily interested in what has been referred to as "cognitive AI" or neuroscience-inspired AI. My group at Tokyo works on a field called "cognitive robotics". The main premise of this research is that intelligence cannot be understood without situating the agent within a body (personally, I think this is an interesting claim, but I am ambivalent about it's stronger interpretations). This research was primarily inspired by Rodney Brooks. I just finished an MSc on discrete planning, which basically convinced me that GOFAI is not the way to go. I now think the main part of the Answer has to do with neural self-organization. So, I am interested in the information processing properties of self-organizing systems. On the topic of good knowledge to have, here are some people and ideas I find interesting/inspiring: Linsker (properties of neural systems self-organizing using Hebb's rule)Hinton/Dayan (unsupervised neural networks)Vapnik (statistical learning theory)Brooks (intelligence without reason/representation)Mumford/Grenander (pattern theory, in particular Lee-Mumford "Hierarchical Bayesian inference in Visual Cortex")Viola/Jones (robust object recognition)Kaplan/Oudeyer/Steels (autotelic principle/progress drive)Hawkins (neural prediction, universal principles of learning) I wonder what other people think of the above list? My question: Following Ben's comment about his colleague who did his PhD while working for Novamente, can anyone formulate a specific question related to FAI that could plausibly be answered by a PhD thesis? In other words, what do people think are the well-defined problems (I am specifically interested in questions relating to the above topics, but feel that precise questions are generally Good Things)-Dan Burfoot> Date: Wed, 6 Dec 2006 23:36:26 -0500> From: ben@goertzel.org> To: sl4@sl4.org> Subject: Re: Re: Re: What are useful for a phd?> > > A build-your-own PhD may result in an advisor who doesn't understand> > or care about your dissertation, and thus has no way of deciding when> > you are "done".> >> > Find a professor who has a grant, and a problem for you to work on.> > Work on it and graduate.> > Better yet, get an MS, get a job, be given a real problem to work on,> > and use it for your dissertation.> > Phil, I feel your advice is overly cynical!!!> > For instance, Moshe Looks (who works with me on Novamente), is just> now wrapping up his PhD at Washington University (in CS). What he> did was:> > -- fund his PhD via working on a professor's research grant, i.e.> working in their lab part of each week on some narrow-AI research> > -- do a PhD thesis on a topic of his own choice, closely tied to the> Novamente AGI system but also with more academically acceptable> narrow-AI aspects (see www. metacog.org)> > In terms of coursework, he took mostly CS courses as expected but also> got permission to do some cog sci courses in the psychology> department.> > His adviser is not extremely expert in his thesis area but is a> lateral thinker, and was happy to supervise his thesis anyway. His> thesis committee includes a number of outsiders including myself and a> couple folks from universities other than Wash. U.> > The point is, there actually is some flexibility in the graduate> education system, if you find the right people and convince them you> are capable of independent work and learning.> > I am not in love with contemporary academia by any means, but Phil,> you seem to be overstating the case against it ;-)> > -- Ben G
_________________________________________________________________
Use Messenger to talk to your IM friends, even those on Yahoo!
http://ideas.live.com/programpage.aspx?versionId=7adb59de-a857-45ba-81cc-685ee3e858fe



This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:57 MDT