9:25 AM: A fascinating plenary lecture by Sylvain Martel on nanorobots capable of communicating with the outside world. He's arguing in favor of combining biological "nanodevices" with electronic intelligence. Some interesting ideas on "motors" formed out of bacterial flagella, using magnetotaxis (following an external magnetic field with an onboard magnetite sensor). He argues that this is a communication channel, since one device can influence its neighbor's magnetic field. Magnetic fields are detectable with external equipment. However, devices may need to cooperate in order to communicate with the outside.
2:09 PM: Neil Gershenfeld's plenary has just begun. "Computer science is one of the worst things ever to happen to computing or science." His problem seems to be with abstraction. I guess the massive global information technological infrastructure isn't good enough for Professor Gershenfeld. At this point I'm inclined to stop listening, as I usually am when somebody makes a ridiculous statement to attract attention. He then contradicts himself by referring to a result of computer science to show a linear-time sort.
2:32 PM: Claim: Locally enforcing constraints leads to global convergence ... ? This seems to be what he is saying, while arguing that asynchronous behavior is irrelevant. It's an incredible claim as it implies that his model ends race conditions forever. He's trying to demonstrate examples of this, but his diagrams are tiny and incomprehensible.
2:53 PM: He starts making more sense talking about problem relaxation to message-passing algorithms. This I can buy: you can relax hard problems to easier problems and solve them much more easily.
3:01 PM: In the end it was an occasionally intriguing but more often frustrating talk. Lots of huge claims and grandiose statements but little clear promise that it will lead to a computing paradigm that will be any more successful than what exists today.
No comments:
Post a Comment