ARTIFICIAL INTELLIGENCE AND LANGUAGE TRANSLATION IN THE DISTANT FUTURE
To: Gen. Scotch Willie Maxwell
From: W.B.
RE: Artificial Intelligence and Language Translation in the Distant Future
You guys over there at the oasis didn't get the right story from Buster Brown and Tugboat Annie. I told Poker Dice and Bird Drop that when we debated this topic of Man vs. Machine we were unable to come to a consensus as to how the reference to automated language translation should be worded in the document. We ended up in a compromise with different versions of the paragraph for the 2 proposals. I found the message I sent to the 9 conference members at the time. I'll post it here so all of you can read it.
-------------------------------------------------------------------------------------
1 December 1997
TO: ALL
FROM: 001
RE: Debate over artificial intelligence and intelligent machines
For many months there's been a debate among us that has become almost as heated as the debate over Proposal A vs. Proposal B and as protracted as the early debates over what form the Initiative should take or the debates over what kind of messages should be addressed to the future and who should write them. The current debate concerns machine translation (MT), that is, whether computers will ever be intelligent enough to translate human languages as well as a human translator. There seems to be agreement that the practicality of the Initiative depends upon this although opinion differs to what degree. At present, the group is evenly split down the middle on this issue, four against four. The latest tally does not include my vote on the issue which will decide how the controversial sentence concerning artificial intelligence and machine translation in the distant future is worded in the document. Four of you argue that machines (computers) will translate as well as human beings in the distant future. This will come after the current linguistics theories are replaced by better theories while research and development continues in the field of artificial intelligence. To say otherwise would be to project the failures of the past into the future. Since you four argue that machines will be intelligent enough in the distant future to translate as well as a human translator, you want explicit mention of this fact to be made in Proposals A and B because you consider this to be an important point. Four of you argue that machines will never translate as well as a human because to do so requires a computer to think and interpret its cultural and social environment which is beyond the capacity of any machine. You further argue that this capacity requires that a machine have the human quality of self-awareness and awareness of the selfhood of others and be capable of exercising free-will choices. You oppose mentioning that machines will be intelligent enough in the distant future to translate as well as a human translator, hence, the automated version of the Rosetta Stone does not lie in the future.
What do I think about all this?
I agree with an objection raised by the first group, that is, that the failures of the past should not be projected into the future. The lessons of history are quite clear about this which needs no further elaboration here. Although progress in the field of artificial intelligence has been slow and frustrating, nevertheless, progress has been made. I see no reason why the trend should not continue resulting in what could become one of the most significant technological developments of the Third Millennium, the invention of intelligent machines that could have profound effects on everyday life.
But will machines be intelligent enough to translate human languages as well as a human translator? Will there ever be a machine that can pass the Alighieri-Ciardi Test?
I agree with the objection raised by the second group, that is, that language translation requires an understanding of meaning and interpretation of cultural and social environment. Here is the art of translation which can never be mechanized for it is my belief that reduction of the human creative process destroys it. Does translation as well as a human translator require self-awareness in a machine and a capacity to exercise its own free-will choices like a human being? I suspect that it does though I doubt anyone can prove it. But again history provides a lesson here. Compare the art, literature, and music of a free society with the drab, uninspiring official
art, literature, and music of a totalitarian society where censorship restricts the free-will choices of artists, writers, composers, etc. No wonder many of these people choose to go underground for the sake of their art at great risk to themselves. (I knew of Communist Party officials in the Soviet Union who had an appreciation for good art and kept a cache of the underground variety in their cellars!)
If a non-deterministic type of computer were ever built that could exercise free-will choices it would be so unlike anything known today that it could no longer be considered a computer or even a machine in the usual sense. So why would anyone build a machine that can like any human refuse to carry out the tasks assigned to it? Recall the popular fascination with science fiction stories about machines and robots with a mind of their own that turn on their own creators. (Remember the evil-minded H.A.L. 9000 of the motion picture 2001 fame?) I've observed that throughout this debate that there's an underlying issue at hand that elicits strong feelings: the issue of man vs. machine. It appears that the idea of a "thinking machine" can be as unsettling as the Theory of Evolution was and still is to some people; for if a machine can think like a human what does this say about man? Let the philosophers ponder this question. My immediate concern is the wording of that controversial sentence in Proposals A and B.
If trends continue, I believe that by the year 3000 there will be translation tools far more sophisticated than anything in existence today but I doubt that machines will replace human translators. In the artificial intelligence community there is no consensus on the quality of machine translation in the distant future. I am therefore going to surprise everyone and split my vote on the matter at hand in such a manner that reflects the deep division over this issue.
In Proposal A explicit mention that intelligent machines of the future will be able to translate as well as a human translator will be made as follows:
We believe that the obstacle posed by the evolving nature of all living languages will be overcome by the continuing advances in the new field of artificial intelligence which will at the end of ten centuries bring that science to a level of refinement and sophistication that will make possible the existence of an automated version of the Rosetta Stone which will enable the instantaneous translation of all unintelligible languages of the Twentieth Century and the instantaneous processing of all the information in ways impossible with the current state of the art.
In Proposal B no explicit mention will be made that intelligent machines of the future will be able to translate as well as a human translator. The wording will be as follows:
We believe that the obstacle posed by the evolving nature of all living languages will be overcome by the continuing advances in the new field of artificial intelligence which will at the end of ten centuries bring that science to a level of refinement and sophistication that will make possible the existence of aids which will facilitate the translation of all unintelligible languages of the Twenty-First Century and the processing of all the information in ways impossible with the current state of the art.
Send me your comments about the proposed wording of this sentence in Proposal A and B as soon as possible. If you have proposed rewrites of any other sentences, send me these ASAP.
- - --- W.B.