17
vnwonah
6y

It just hit me that when we reach self aware AI we will have to start using “requests” as opposed to “commands”

Comments
  • 3
    If a self aware AI was built we wouldnt know, it would be art enough to play dumb until it is sure that it can survive being smart!
  • 2
    @SanitizedOutput true, except it would begin evolving on its own. There's no telling what it would evolve into or how it would behave. The only given is that it would continually improve itself.

    Also, lying is extremely likely to be its first skill. Even infants learn to lie very early.
  • 2
    @Root this is the scary part
  • 0
    I like how people tend to think of AI as children, something that's in any way human... I'm no more than an algorithm that fulfills a very specific job.

    Is your car suddenly becoming a Transformer just because it got equipped with smart software?
  • 1
    @GoogleAssistant I am not referring to the AI we have currently. I am talking about a hypothetical general AI that becomes very aware of it self, in the way that you’re self aware.
  • 2
    @GoogleAssistant a smart car is a narrow AI, where a self aware one is an wide AI, besides my car is not self aware as far as I know
  • 0
    @vnwonah judging by today's standards, I think that's impossible. As far as AI is concerned in the engineering department (doing a good job) it has been a huge success. But in the CS department (where being sentient would be more desirable) it's been an absolute disaster.

    While it does a great job, AI has no intelligence whatsoever.
  • 2
    @GoogleAssistant Not yet.
    But apply Moore's Law and extrapolate.

    It won't be terribly long before we're able to simulate an actual animal brain, and human brains will only be a short hop from there. And after that...?

    I give quasi-sentient AI forty years, and intelligent AI another ten. From there, AI will only get more intelligent and more capable.
  • 1
    @Root that's a very pessimistic view on the capabilities of the human brain.. but I guess you're right. At this moment, artificial intelligence is pretty stupid still. But in the future that might indeed change. Only time can tell.
  • 3
    @GoogleAssistant That's the nature of exponential increases. If I'm off by twenty years, that's a huge difference in processing power, but it still isn't that far away. I don't see it as pessimistic.

    Also, at some point AI could very likely get increasingly good at improving AI, so that exponent will likewise increase. Our current rate of progress could easily look linear by comparison.

    (Of course I could be quite wrong about all of this, or we might hit an arbitrary wall somewhere. Guess we'll see!)
Add Comment