Saturday, June 15, 2013

GF 2045: Section 2

My battery indeed ran out and prevented live blogging of the second session, so I'm going to give a very quick summary.  This section concerned the current technology of android and human-computer interface.

The highlight was Hiroshi Ishiguro, a Japanese robotics professor who builds life-like androids.  He's a warm, funny, appealing speaker, and brought a couple of androids with him, including a replica of himself.  The state of the art in androids is that they look fairly life-like but behave like stiff, indeed severely autistic, individuals with computer generated voices saying pre-determined things (although a video of an android and a robot talking while both running cleverbot was fairly hilarious).

This has been Ishiguro's life work, and he describes his motivation as being to understand what it means to be human - only by trying to build simulations can we understand what really makes humans tick.  I have some sympathy with that motivation.

The other thing that was very helpful was a pair of professors from Berkeley (Jose Carmena and Michel Maharbiz) who gave a talk about the current state of Brain-Machine interface.  It's pretty primitive - they can stick a grid of a few dozen needles into the brain and read stuff out of it, but it will only last for a year or two.  In using robotic prosthetics, they currently cannot get anywhere near the needed degrees of freedom to control all that the arm can do.  Plus the requirement to get wires out through the skull is a major problem (infection risk).  They are currently working on an approach ("neural dust"), in which a bunch of tiny disattached receptors (the dust) would be scattered in the cortex and communicate via ultrasound with a base station at the surface of the cortex, which in turn would relay the information wirelessly to another station attached to the skull exterior.  If they can get it to work, that sounds like it would be a big advance.

In short, this stuff is currently light-years from being able to control anything like a complete avatar.  Of course, there's still three decades between now and 2045.

Still, I have to say that the ability of robots/algorithms to displace humans from jobs is many decades ahead of the ability to provide much improvement to the human body.  My concern is that it's likely to stay that way.

No comments: