Went to see a lecture by Prof Kevin Warwick at the Lit and Phil few weeks ago (he’s the guy who talks about cyborgs and AI who makes headlines by getting chips inserted). A very good lecture; I’d been afraid of pretension but he was not only very down-to earth and smart but he had a great sense of humour into the bargain. It was slightly marred when someone’s mobile disgracefully went off during the questions (a new phone and I hadn’t worked out which profile was the silent one; terrible shame is mine).
I’m not sure whether the machines are going to be taking over any time soon (I should probably do a post about that sometime…) In fact, he avoided that topic. But he did show the scipts of several Turing tests near the end (NB in a five minute test, you only get about a single OHP of conversation; not much). He asked people to put up their hands as to whether they thought it was a human or machine.
Interestingly, most people tended to always say “machine.” I suppose there is a big bias to “I never want to get fooled into thinking a machine is a human; I don’t care if I keep getting it wrong the other way around.” But a lot of people (including myself) got fooled by a program that cracked a joke about Monty Python. Looking back, the joke came rather out of nowhere, or from a trigger that a programmer could have fairly obviously spotted.
But it did suggest some new essential rules for Turing tests:
1) The human has to be articulate. If a machine just responded to everything with “whatever” then it would perfectly imitate a type of teenager. More seriously, imitating a teenager not conversing in their native language (which made recent headlines) is shifting the goalposts.
2) The human has to be actually trying. Some people in the examples we saw were deliberately imitating a machine, or just being awkward.
…and to be fair, two to help the machine:
3) You cannot ask the machine things like “what does it feel like to sneeze?” or “what was the canteen here serving?” that depend on having a physical body
4) You should not interrogate in depth about biographical details (“What was the name of your first schoolteacher?”) as that would add a second layer of necessary deception; it takes training for a spy to pull off that trick.
The question I wanted to ask, but couldn’t because of limited time, was “what does it feel like to control something from the implant in your optic nerve? How do you do it?” As a writer, that would be most useful to me; I am on my second book featuring someone with computer implants they control internally, and I’d like it to be as plausible as possible.