Ran out after Kate Hayles and dashed into another panel on artificial intelligence and chatbots, which is turning out to be really good.
Keep it Simiple (and female?) A reaction to the masculine robot or HAL from 2001 who came to destroy our world – easier to relate to a non-threatening young administrative assistent style female chat bot.
Tim Menzies:
Computers are already smarter than us, but using methods so alien that we’ll never catch them up, and they may already be beyond our understanding. (planes don’t fly the same way birds do but work, in some ways, better – certainly differently)
Limits to AI (these are the three objections to AI everyone has to point out
– Godels incompleteness theorem
– Suchman/Clancey situation activity/cognition (what people SAY they do isn’t necessarily what they really do so not possible to design robots)
– Cook’s NP-completeness
E.g. John Koza, Keane and Matthew – a program that can take any human-developed patent and design a NEW circuit that will do the same thing, differently. They have thousands of examples.
AI can reason about larger theories than humans. Menzies et. al. 1985 – a simple AI system outperformed the human who’d encoded it. Why? short-term memory just not big enough in humans.
Being able to do something doesn’t mean you can explain how you do it or what you’ve done.
Often AI systems where it’s really NOT obvious how they’re working something out. THey’re opaque, alien to us.
Is this worrying? No. we can share – we bring the power, they’ll help us out now and again, and maybe sometimes we’ll understand what they’re trying to tell us. Worrying about AI is like being the spoilt only child worried about the birth of a sibling.
Helpful benign AI agents eager to share all they know with you. Yet it can’t. Sometimes AI tries to talk to us, and we just say “–wha?”
Machines reading and listening
Surveillance cameras aren’t really working. E.g. delivery man in NY who was trapped in a list for three days despite security cameras and people searching for him – either he was in the camera’s blind spot (!?) or the security people were just not looking at the cameras. Other examples too of missing cameras.
Machines reading.
Claus
> Computers are already smarter than us
What does “smarter” actually mean? What does “intelligence” mean? Machines (or computer programs) are (yet) not able to *think* at all. “Thinking” would mean that they are aware of themselves–which is not the case right now (2005). Programs like the classic LIZA are mereley attempts to emulate human reaction. Until advanced hardware has been developed with the ability of self-organisation, I would not dare using the term “intelligence” (or “A.I.” for that matter) at all when related to machines.
Claus
P.S.:
> LIZA
I think that the name was actually ELIZA (presumably after “My Fair Lady”). This program was written by Joseph Weizenbaum, and its task was to “interact” with a human by “answering” questions with one out of a set of predefined sentences, thus giving the impression that a human being was “on the other side” instead of a machine.