while i had a job at the gym, i would see prof. sapolsky work out. i always wondered what he was thinking about when he did pull-ups, given his special insights.

Well, of course, it is tempting to go for something like, “That the wheel, agriculture, and the Macarena were all actually invented by yetis.” Or to do the sophomoric pseudo-ironic logic twist of, “That every truth can eventually be proven.” Or to get up my hackles, draw up to my full height and intone, “Sir, we scientists believe in nothing that cannot be proven by the whetstone of science, verily our faith is our lack of faith,” and then go off in a lab coat and a huff.

Sapolsky’s books are also great, highly recommended, esp. A Primate’s Memoir.

re: Minsky via Yiming, I saw a decent, if low-budget, series of short films called “Robot Stories” @ the Asian Film Festival a few years back. One of the more compelling shorts is about a couple who’s been backing their consciousnesses up for centuries, to guard against the physical failures of their mortal bodies. do we need to feel a sense of loss to be fully human? will we ever have a high-enough resolution recording device that will render loss obsolete? what cannot be recorded and “memorized?”

i was just recently exposed to the edge annual questions when i picked up this book: http://www.amazon.com/What-Believe-but-Cannot-Prove/dp/0060841818

“What Do You Believe, But Cannot Prove?”

highly recommended, i’ll loan it to anyone who wants it.

Ah, then there are the obligatory AI pipedrea…ahem… visions in there. Marvin Minsky, a celebrity AI academic if there ever was one, describes immortality by cybernetics, downloading/backing up the information contents of human brains into machines and augmenting failure-prone organics with mechanical parts. Going along with the semi-fantasy, a direct mind-machine information interface would go a long way toward alleviating the lossy nature of human-computer interaction. A request to the machine to “do what I mean” could be made far more literally.

Jordan Pollack starts with your basic intelligent agent that learns from user feedback and concludes with digital sentience through self-evolving software, slanting decidedly toward the non-symbolic approach to AI. Seeing that Skynet hasn’t arisen from the orderly and carefully crafted databases of Cyc, perhaps there’s something to the non-logical, throw-everything-at-it-and-make-it-learn-somehow approach. I just don’t know what, besides a lot of hand-waving.

Terrence Sejnowski proclaims that a breakthrough in understanding intelligence is “just around the corner.” We’ve been hearing this for a while now, though. That corner seems rather impregnable so far. Statistical learning algorithms only seem to go so far.