Notacon submission
Sean has agreed to co-present with me at Notacon, which will be good. I think between us we'll do a better job of holding an audience's attention for 30-60 minutes than I would have been able to do alone. We submitted our proposal yesterday evening; here is the talk outline:
"Walking before we run: how biology is inspiring progress in AI"Computers may be outstanding tools for processing data, but they still require constant human oversight and perform poorly at many tasks that come naturally to even ‘lower’ animals.   Meanwhile our attempts at making computers do these sorts of tasks are giving biologists and psychologists new insight into how natural systems work.
This talk will show you examples of recent work in artificial intelligence that draws its inspiration from biology, and life sciences research that is informed by computer models. We hope to convince you that these are useful for everything from spam filtering to Mars rovers to diagnosing heart failure.Â
Suggestions for content are welcome. At this stage I have only a pretty vague idea of which examples to give and how to tie it together as a whole.
update: the proposal's been officially accepted.
Comments
Interesting... do you have some more juicy information on such a topic. I am assuming it is sufficiently different than normal Bayesian techniques.
The general principle is that neural networks, genetic algorithms and artificial immune systems make good classifiers for fuzzy data sets, like the spam-ham discrimination, and are good at coping with datasets that change over time, as with spammers' evolving strategies for defeating the filters. I must confess that the specific example of a spam filter was something I threw in there without a great deal of hard information to back it up, but there is some interesting work out there, and here are the few pointers I have to hand (with a little google work to back up my statements):
It can be difficult finding any sort of detailed information about what techniques a given commercially available filter uses and how, partly because it's vauable IP and partly because they don't want to help spammers defeat their product, but hopefully the information I have is of some help.
Incidentally, if you can excuse the spam, have you looked at the Notacon speakers list? I reckon a talk about some of the things you blog about—particularly on the uses of blogging and social software—would be of interest to a good number of the people there.
Hmmm... I could give a talk about blogging. I would want it to be more specific that just "general blogging." Maybe about "Uses of Blogs in Higher Ed." or "Uses of Blogs in the IT industry of Higher Ed."
I agree. I think the more specific the talk is, the more likely it is to be of interest. Most of the audience there would be familiar with blogging—a fairly large proportion probably blog themselves—but I think the benefits to an organisation like ITS of running a blog system are less obvious, so more worth talking about.
i've actually been talking about using bioinformatics as a means to do spam or general payload analysis for a while. i gave a talk at cwru's ACM meeting in january on this topic:
http://monkey.org/~jose/presentations/cwru-acm05.d/
there's a lot to be learned from the life sciences, since they have so much data to sift through. i'm curious to see eldan's talk on the subject, since he's one bright guy and sure to bring up useful information and insights.
Thanks Jose :)
The vote of confidence is much appreciated. However, I should probably explain that the sorts of things we'll be talking about don't have a great deal in common with bioinformatics, because the bioinformatics people tend to use fairly classical computer science techniques—the sorts of things I had to learn for the Algorithms half of my PhD qualifiers—whereas our focus is on techniques that are themselves inspired by biology.