" ... Fear of enslavement by our creation is an old fear, and a literary tritism. But I fear something worse and much more likely -- that sometime after 2020 our machines will become intelligent, evolve rapidly and end up treating us as pets. We can at least take comfort that there is one worse fate -- becoming food -- that is mercifully highly unlikely ..."Paul Saffo, director of the Institute for the Future, on the evolution of smart machines in the 12/06 Communications of the ACM.
Paul, who we have connected with since the late 70s, as a practitioner, I disagree here. I don't think the type of intelligence he suggests will be here sometime shortly after 2020, if ever. There are still very fundamental problems with cracking the problem of intelligence. Its not about faster machines, or more memory, or more connected networks .. but about understanding how to make things smart. Its not about knocking down straw men like playing chess, but its about very deep forms of pattern recognition that allow us to abstractly understand our world. The most common computing intelligence we use today is still simple search. I think the old question of whether a dog could make a machine smarter than itself has still not been shown to not apply to us as well.
No comments:
Post a Comment