Linked by Eugenia Loli on Sat 23rd Jun 2012 20:18 UTC
Hardware, Embedded Systems Artificial intelligence is still a long way from delivering the human intelligence in robot form that has long been common in science fiction.
Permalink for comment 524607
To read all comments associated with this story, please click here.
RE[2]: Tron
by zima on Sat 30th Jun 2012 23:19 UTC in reply to "RE: Tron"
zima
Member since:
2005-07-06

It seems there's a strong tendency to shift goalposts in the field of AI.

Like zima already said, there's a risk of setting the bar so high as to rule out animals and humans. If we're to be objective, our litmus tests shouldn't focus around humans proficiencies but instead be inclusive of any intelligent life in the universe.

Here's a challenge: come up with a satisfactory litmus test that animals and humans can pass but ultimately computers cannot.

Conversely, and most tellingly: if we would be able to demonstrate even quite basic AI (like the fairly old chess programs that I mention in the previous discussion about AI, linked in my 1st comment in this one) to the people few generations back, I'm fairly certain they would be mighty impressed (...at least, at first), if not outright suspecting divine influence (vide cargo cults). I believe chess was held to be a good test of AI 50 years ago...

It even has a term: http://en.wikipedia.org/wiki/AI_effect

People seem to expect from AI to be at least on par with the best humans, in all "higher" fields. While... to be worth it, it needs only to be better than an average human practising only its one particular discipline or activity (AI flying airplanes doesn't have to now anything about poetry, for example; it even shouldn't have too much in common with AI flying highly manoeuvrable, highly evasive / survivable cruise missiles, beyond similar "understanding" of aerodynamics)

There's one perhaps even more curious term http://en.wikipedia.org/wiki/Moravec's_paradox ...it seems that "animal traits" are harder.

Reply Parent Score: 2