Discussions About Artificial Intelligence

Sometimes it is interesting to think about thought itself – metacognition. If one considers all the information we have about how thought works, it's even more interesting to consider that something called artificial intelligence is being discussed – if we're not entirely sure about all the details of our own, of human, thoughts, how should we know how to give such abilities to a machine? The principle is simple enough: impart on a machine the ability to think, to perceive what is around it and to adapt – to make choices that will “maximize chances of success” (Russell).

Since 1956, when the term was coined by John McCarthy (Association), artificial intelligence has been the subject of both many legitimate scientific endeavors as well as a huge number of science fiction books and movies. In the former case, progress has been made toward AI, but machines are still incapable of making procedural, dynamic choices in relation to their environment; the criteria for creating true AI is daunting, to say the least. In the latter area, artificial intelligence has been portrayed in a wide spectrum. Depending on the work, it is either the downfall of humanity or our salvation – or even somewhere in between. Many creative minds are fascinated to speculate about how thinking machines could affect our way of life. In both cases, the ability for machines to truly think is an idea of the future, an idea that becomes evermore clear due to ongoing research.

However, I want to talk about one area of artificial intelligence in particular: the ability for a machine to communicate with humans. That doesn't seem too complicated, does it? Machines are capable of communication with humans already, right? Alan Turing, an English computer scientist, speculated that, if a machine could fool a human into thinking he or she was actually corresponding with a real person, that machine could be considered intelligent (Turing). This is obviously a lot trickier than a computer spitting information back at you – this is a level of interaction that is indistinguishable from real human-to-human communication. Want to know what's the most surprising about this theory? Turing's article describing this idea was published in 1950. Archaic computers were around at this point, but the term “artificial intelligence” was still six years away.

Enough about the history of the terminology, though. What implications did Turing's ideas have on what scientists are doing today? It turns out there's actually an annual competition (which happens to offer cash prizes) for programs entered to compete in the “Turing test.” As described by Turing, this test involves a judge and two correspondents. One of the correspondents is human, and the other is a machine. None of the participants are visible to one another. The object of the test is for the judge to exchange messages (written only – think instant messaging) with each of the participants in turn for a limited period of time. At the end of the round, the judge has to identify the human entity. A machine is said to have passed the test if it can consistently fool these judges.

The aforementioned contest, known as the Loebner Prize, offers $100,000 (as well as a medal) to anyone who can create “the first computer whose responses were indistinguishable from a human's” (Loebner). If the grand prize isn't a good enough incentive, first place competitors receive $3,000 anyway (that is, if they win the competition but haven't necessarily developed a computer that can consistently fool all of the judges). Needless to say, this particular instantiation of the Turing test has drawn competitors since its inception in 1991. No one has claimed the grand prize yet, but results of the most recent contests have been recorded and the transcripts can be downloaded from Loebner's website: http://www.loebner.net/Prizef/loebner-prize.html

The implications for having a machine that can realistically converse with humans are definitely vast. You need only think about it for a moment – each computer would be able to interface with its users much more seamlessly. Further, people wouldn't be able to tell the difference between human and computer entities in social environments like chatrooms. These things don't seem all that revolutionary, do they? Consider this: a machine that could emulate human conversation would be able to adapt to behaviors. If this were to be applied to applications, it would ostensibly increase productivity by a significant amount – if the program is familiar with your workflow and habits, it would be able to genuinely help you along (and I don't mean that weird paperclip guy from the old Microsoft applications).

This raises the question I want the class to think about: If we did create a computer that was able to communicate just like a real human, will we have, as Turing predicted, discovered true artificial intelligence? Or will we merely have a huge pile of scripts and algorithms that don't really do anything dynamic, but can fake conversation well enough? Do you think that artificial intelligence will require a more revolutionary or transcendental step in computer science, or do you think it's attainable in the direction we're going now? Regardless of the overall impact AI might have on humanity, a scientific revolution (in the digital field in particular) that brings thinking machines into existence would likely have other unforeseen but magnificent effects as well. Hopefully I gave everyone a bit to think about – I look forward to a class discussion. Thanks for reading!


Works Cited:

Association for the Advancement of Artificial Intelligence. Web. 07 Oct. 2009. .

"Loebner Prize for Artificial Intelligence." Hugh Loebner's Home Page. Web. 07 Oct. 2009. .

Russell, Stuart J. Artificial intelligence a modern approach. Upper Saddle River, N.J: Prentice Hall/Pearson Education, 2003.

Turing, Alan. "Computing Machinery and Intelligence." Mind LIX.236 (1950): 433-60. Oxford Journals. Web. 7 Oct. 2009. .