An argument against the possibility of super-intelligence is that the intelligence of a creation will be limited by the intelligence of its creator. How reasonable is this argument?
$\begingroup$
$\endgroup$
Add a comment
|
3 Answers
$\begingroup$
$\endgroup$
4
Of course the intelligence of a product is limited by the intelligence of its creator. Just not to the intelligence of its creator.
That would be about as reasonable as the idea that the speed of a car is limited to the speed of its creator.
Or the playing strength of a chess program to the Elo of its creator.
Or the ability of a neural network to differentiate between dozens of dog breeds to the dog expertise of its creator.
So, not very reasonable.
answered Mar 2, 2017 at 14:07
-
1$\begingroup$ On the other hand, the argument is reasonable since it claims the existence of a super-intelligent machine is dependent on us knowing how to build it (so if we don't know how to build one and we don't accidentally build one, super-intelligent entities will not exist). Currently, humans are not smart enough to build a super-intelligent algorithm, and we have no idea how to build such an algorithm. $\endgroup$ Commented Mar 2, 2017 at 14:11
-
$\begingroup$ @TariqAli ,But what if it's happening right now.What would you say? $\endgroup$ Commented Mar 2, 2017 at 17:24
-
$\begingroup$ Excellent answer! @TariqAli I think the real problem won't be if the machines are smarter than us, but if they're not smart enough and become antagonistic as opposed to cooperative, destructive instead of equilibrium seeking. The dumb machine threat is the "grey goo" model, but powerful, AI, limited by, and exploited for, selfish reasons by "un-evolved" humans is probably the most realistic "next threat". Imagine a world where a single financial firm owns everything because they have the most powerful algorithms (as opposed to companies who care about the algorithms more than the money.) $\endgroup$– DukeZhouCommented Mar 2, 2017 at 22:17
-
$\begingroup$ @TariqAli - I watched a video where they were talking with the developers of the XBOX Kinect. All the developers did was create the neural net software for it. From there, the neural net figured out what to do based on the training data. The developers said they would never have been able to get it to work by programming it themselves. So I'd contend that this is an example of an AI that figured out how to build something that humans probably couldn't have done, at least in a reasonable time-frame. $\endgroup$– DunkCommented Mar 8, 2017 at 20:01
$\begingroup$
$\endgroup$
2
AI is frequently used to discover things that would take laborious amounts of time for humans to do. For example, AI can be used to find the optimal configuration for a mother-board layout, or identify best fit parameters for a financial model. Frequently, the AI can do a better job at a task and do it more qucikly than a human. Therefore, in many applications, the AI is already more intelligent than the creator at specific tasks.
Here are just a few things that AI can already do better than humans:
- Playing Chess
- Playing Jeopardy
- Detecting Cancer
To argue against the possibility of a super-intelligent AI is somewhat of a moot point since it has already been proven.
$\begingroup$
$\endgroup$
An important difference between a human intelligence and an artificial intelligence is that the artificial intelligence would "think" much faster, perhaps millions of times faster than our brains do. Our neurons are slower than transistors. It also wouldn't be limited by evolutionary biology and could presumably modify itself or create a replacement AI that operates more efficiently. Do this modification an arbitrary number of times and it becomes super intelligent.
