The Future of Artificial Intelligence
Terminator 3 Poster, showing AI robots (Source: IMDb)
It's difficult to know precisely where AI will go in the future - though the ultimate goal is to produce a human intelligence, there are a number of routes that AI could end up taking:
- The "SkyNet" route, which is a staple of science fiction,
where we do end up with artificially intelligent computers, but
that these computers then judge the human race to be a threat to
their survival, and focus on the removal of this threat to
safeguard there own future - essentially, we'd be creating the
very thing that makes us go extinct, echoing evolution. With
that being said, science fiction has also come up with a
solution to that problem - Isaac Asimov's "Three Laws of
Robotics", which any right minded person would build into a
smart computer to provide a safeguard against it deciding that
we are a threat that should be eliminated.
This is the viewpoint shown by William Joy in his 2000 article for wired magazine (Joy, 2000) and Hugo de Garis in the following quote “I see the debate over whether we should build these artificial intellects as becoming the dominant political question of the century,” (de Garis, 2005)
- "The Singularity" by Vernon Vinge (Vinge, 1993), which
argues that the acceleration of technological progress had led
to “the edge of change comparable to the rise of human life on
Earth.” In essence, this thesis argues that there will come a
point where the abilty to corelate information and come to a
correct conclusion based on that information with happen at a
faster speed than we can think of the questions, essentially
leading us to a point where all the knowledge of the universe is
immediately available to us through our computers. One of the
main requirements for this singularity ot happen however is that
computer power continues to increase as per Moore's Law, and
also that we develop an artificial intelligence that is capable
of winnowing the facts from the mass of information that we
would be generating every instant.
- Another option is that AI will develop along non-human lines. As it grows and develops its own way of looking at the world and builds its own neural networks, an AI that was originally based on the human brain, would diverge from this, as it evolves on its own. With this possibility, our first meeting with alien life may very well not be with E.T. but instead with something of our own creation that has grown to look at the world in which it exists in a manner in which we might never be able to understand.
In practical terms however, we can, in all likelyhood look forward to the following things to come out of AI development in the next few years:
- Self driving cars.
- Automated housework.
- Automatic food re-ordering.
- Automatic bill payment.
- Smart "agents" that will see out information and present it to you based on partial sets of data.
As to how these marvelous things are going to actually be implements, they rely on a number of techniques and methods, described simply below.
Neural Network(s) is software that is based around the way that an actual brain works. Neural networks are designed to recognize patterns in data and predict an output from a given set of information. A classic example is the use of neural nets to predict what stocks are considered to be undervalued given all of the information about the stock at that moment in time (P/E Ratio, Historical Fluctuations, Sector Performance, etc). Neural networks require training before they can predict or learn.
Mobile Agent System(s) - This is software that does something on the user’s behalf. Essentially it's a mini-you that you can send off to do something, and it will do it if it is within the parameters that you have set it, for example, if you want to go and buy a car and you know what model you want. A more simple (and current) type of agent would be something like insurance comparison sites.
Genetic Algorithms - A method of machine learning, which copies its method of creation from nature. It does this by looking at the way DNA changes based on differing circumstances and population size, and using this to develop a computer program, producing the best able to survive from the "gene pool".
Vinge, Vernon - Vernon Vinge on the Singularity [internet] 1993. Available at: http://mindstalk.net/vinge/vinge-sing.html [Accessed 13 January 2010]
Joy, William - Why the future doesn't need us. [internet] 2000. Available at: http://www.wired.com/wired/archive/8.04/joy_pr.html [Accessed 13 January 2010]
de Garis, Hugo - The Artilect War: Cosmists Vs. Terrans: A Bitter Controversy Concerning Whether Humanity Should Build Godlike Massively Intelligent Machines: ETC Publications