First post, by badmojo
- Rank
- l33t
I saw the movie ‘Her’ the other night and it was surprisingly good. Like all movies I can think of that revolve around an A.I character, the conflict was generated by the question of the A.I’s ability to have emotions and individual thought. It got me thinking about the discrepancy between science fiction and reality, and while I can understand why the idea of a true emulation of a human personality is so compelling – I’d love to have a meaningful chat with my PC – I don’t think this will ever be possible. A.I to handle specific tasks is progressing, for example driving a car, which is relatively complicated but is a task with finite inputs and outputs. In fact I think a well tested A.I could do it better than humans, who are easily distracted and often act impulsively and break the rules.
But when it comes to something like creating Sigfrid von Shrink (the A.I therapist from Frederik Pohl's Gateway), regardless of how powerful the hardware is, the software would have to make sense of an infinite and often illogical set of inputs, because the human mind is too complex to predict. Even maintaining a believable conversation is unlikely I think – smoke and mirrors efforts like the 2013 Loebner prize winning chatterbot Mitsuku are a fun diversion, but there’s nothing particularly intelligent about it. It basically just boils down to a massive ‘IF’ statement – a hardcoded set of responses that gets fine tuned to the point of appearing to respond in a realistic way, but you have to try very hard to suspend disbelief. 'But this is what makes it artificial!' you say? I guess so, but this approach is just too simplistic in my mind. You might be able to model an ant's mind in this way - maybe - but never a human. Human's have some extra level of magic going on, and it can't be reproduced with zeros and ones.
If anyone has a more optimistic take on the future of A.I then I’d like to hear it – I’m willing to be convinced!
Life? Don't talk to me about life.