Robots, standup and be counted
Every so often there’s a list categorizing robots, the top ten X robots. Like this fun one that IEEE Spectrum recently did on humanoid robots, highlighting ASIMO and four more recently developed bots.
Alternatively, Heather Knight has been getting some press for the opposite kind of list. Rather than a top-10 she’s doing a robot census at CMU. Frank Tobe of the Robot Report also left some interesting stats in the comments, pointing to current robot usage numbers and projected future usage.
So far she’s tallied ~550 robots on campus, and is getting the word out to count robots more broadly. So log in and count yer bots!
ez hcg | A lot of interesting posts on forex analysis.
Hello!
I just read about your work in the recent Popular Science issue (congrats!) and so found my way to your blog. I was thrilled to discover someone approaching machine learning (for lack of a better, broader term) as you have. I, alas, am not directly involved in robotics or AI research or anything like that, but I stay as current as I can on the field. I’ve long thought that the best approach to programming robots to perform tasks is through a learning process directed by humans (at least at first; naturally we’d want to teach one to teach others….
). It’s not worth too much of course, as I have zero credentials in the field, but I think you are absolutely on the right track and think your work and ideas should get much more publicity and recognition.
If I may ask (just ignore this paragraph if not
): your approach certainly seems well suited for robots intended for a role working with humans, but I was curious if your ideas are being applied more generally? In other words, why couldn’t the basic idea – that computers should learn most anything that isn’t a linear algorithm through a “Q & A” or “Model, Try, & Learn”-type approach – be applied to ALL situations where a machine intelligence needs to learn something that cannot be completely and/or sufficiently specified ahead of time? I could see it working even to build (“groom”? “educate?”) a better Human-Machine interface layer in desktop operating systems (Is that person your Mom? Yes? Want me to remember her email? …. You just rebooted…why? … It’s Friday at 2pm, should I go ahead and close your spreadsheet and open YouTube and look for new shows? etc.). Maybe that’s being a bit too far-fetched, but I think the idea in general is correct. I’m not sure why we humans ever thought we could program a computer to do something we had to learn through trial and error and study and over time via some sort of straightforward set of algorithms/processes.
The advantage software has, it seems to me, is once you’ve taught a particular robot or piece of software to do something, it’s almost no additional effort to replicate that now-experienced, intelligent piece of software/robot countless times. Perhaps, to bring it back to your work, humans teach a robot the basics using your method; so it learns the generic parts of some job: folding laundry, helping the elderly to remember and take meds, etc.; that “version” is then stored and replicated as the “baseline.” Then many “baseline” robots can be produced and when they are assigned to a specific job, it continues to learn the specifics just as you’ve shown is possible and described. Much like when a human worker changes employers for the “same” job; as you’ve said, we KNOW that it isn’t REALLY the same. Not entirely, anyway. There are always local variations, and far too many to account for ahead of time.
Sorry to ramble on so long – it’s just exciting to read about work being done like this. I’m very impressed with all that you’ve already accomplished and look forward to seeing more and more fantastic ideas, success, and robots coming from your work.
Cheers!
Scott
Comment by Scott Stratton | October 2, 2012
Thanks for your comments, Scott. Yes, I think that certainly the notions of Socially Guided Machine Learning should not apply only to robots. That happens to be my particular focus, and I love that embodying these algorithms in physical form opens up so many new channels of communication and more human-like learning interactions. But certainly all of the visions you’re describing are the ones we are working towards! All very exciting. There are a lot of big challenges still, but I think we are chipping away at them and I do think that interactive learning will be one of the things that makes robots a reality sooner rather than later.
Comment by A.L.T. | February 5, 2013