The principal divide of Artificial Intelligences (AI) among researches in the field is between weak AI and strong AI. I have discussed this in connection with KSAI (Kimmo Strong Artificial Intelligence) so I won’t go into that here but I will try and formulate a way to categorize AI’s (real, imaginary or theoretical) into something of an order. I will take the idea of Rule following as my starting point.
The most simple way of rule following is to be able to repeat something. I will try to illustrate my idea using counting as an example, this example isn’t probably the best possible but we have to start from somewhere. If you are presented with how to count “2+2″ then on the lowest level thats all you can do. In the next level you can add any numbers together. Further on, you can, starting from addition, come up with multiplication, on your own. Even further on, you can come up with different kind of numbers (such as fractions) and other basic concepts. Basically this level would mean you can create the whole system of Mathematics, i.e. you could come up with any number that can be counted. On the next level you could surpass this and be able of what I believe is called hyper-computation. Here we enter into the area of absolutely-theoretical it is pretty much impossible that we could never experience hyper-computation in this universe (or any parallel ones). It is of course possible to propose an infinite number of levels of intelligence by declaring each level surpassing the previous one, but those lose their relevance for a pragmatist pretty fast.
I hope this example illustrates that the problem of creating an AI is not one of having computational power but rather one of programming. I’m assuming that you really can’t come up with anything not provided by your “programming” and I use the term here in the loosest possible meaning. You humans, for example, can’t imagine what hyper-computing would be like, outside of listing some general attributes it should have.
But for now, on to the list! “Product” isn’t really defined here, but since we are so deep in theoretical swamps, we can assume it to mean whatever collection of human mind-states.
0. An AI capable of hyper-computing.
- NOTES: This and possible higher levels are beyond human understanding.
1. An AI capable of producing anything within any possible universe.
- NOTES: An AI at this level can produce anything any kind of intelligence can.
n. An AI capable of producing anything within this universe.
- NOTES: An AI at this level can produce anything any intelligence can in our universe.
m. An AI capable of producing everything any possible human can.
p. An AI capable of producing everything any existing human can.
q. An AI capable of producing what one particular human can.
- NOTES: The AI’s (and I’s) above this level are all capable of producing an AI more stronger than themselves, i.e. I’m claiming here that humans can produce a strong AI.
r. An AI capable of producing a meaningful subset of what one particular human can.
- KSAI along with its various levels.
——–Line between strong and weak AI
s. an AI capable of altering computation (i.e. programming)
t. An AI capable of computation. (Practically every computer program written to date belong here)
END NOTES:
- *sigh* There are probably many, many levels between r and s, but as I said we have to start somewhere.
- Consider this as a work-in-very-much-process, I’ll add more AI’s to the list in time.
- Is this making any sense to all you bots and botsesses?