Monday, July 20, 2009

Turing Test

Tyler Cowen and Michelle Dawson have written an interesting paper reviewing Alan Turing's famous 1950 paper in Mind called “Computing Machinery and Intelligence”. Here are some interesting snippets.
Our interpretation fits the broad outlines of Turing’s life. Turing was gay and he was persecuted for this difference in a manner which led to his eventual suicide. In mainstream British society of that time, he proved unable to consistently “pass” for straight. Interestingly, the second paragraph of his paper starts with the question of whether a male or female can pass for a member of the other gender in a typed conversation. The notion of “passing” was of direct personal concern to Turing and surely in more personal settings Turing did not view “passing” as synonymous with actually being a particular way.

It also has been speculated that Turing was autistic or Asperger’s, which suggests his mind was of a very different nature, compared to most of the people he knew. Turing probably was not aware of these neurodevelopmental concepts as such (they had not yet entered standard English-language discourse), but surely he knew, growing up, that he was in some ways very different from others. In public school he was judged to be “ludicrously behind” with “the worst” writing ever encountered, and he was singled out as “bound to be a problem for any school or community.” While we cannot be sure whether Turing was autistic, it is clear from published accounts that other people noticed he thought and acted in highly atypical ways. Again, Turing himself could not pass a test of imitation, namely that of imitating the people he met in mainstream British society, and for most of his life he was acutely aware that he was failing imitation tests in a variety of ways.

...

Turing does not make this point but the reality is that many human beings, intelligent and of mature age, could not pass what we now call a Turing test. This includes many human beings who would do well on IQ tests or other traditional measures of intelligence. Some autistics provide examples, as would other individuals with non-standard neurodevelopmental paths. For instance, an autistic might not comply with the presumed or implied typical social context behind many of the proffered questions. Many answers from autistic individuals might seem “off” to the judging panel, so there is a good chance autistics’ distinctly atypical information processing would fail to pass the imitation or Turing test.

The available evidence is sketchy but we do know that one BBC radio producer doubted Turing’s ability to communicate and complained about the “definite hesitation” in his speech. There is also a transcript of a discussion between Turing and several other scientists; one commentator observed: “Reading the transcript is rather like reading the conversations generated by computers... Few of the discussions can stick to a point or actually address a question!” One of the scientists argues against Turing by implying that Turing’s atypical way of processing information cannot be human.

...

So how does Turing close the paper and what does he see as the best path going forward? In Turing’s penultimate paragraph he presents two paths. The first is to increase the computational abilities of current machines, as we might do by improving their abilities at chess. An alternative is to “provide the machine with the best sense organs that money can buy, and then teach it to understand and speak English. This process could follow the normal teaching of a child.” Is Turing implicitly—and with dry British humor—referring to the “normal teaching” he received as a child, and its unforeseen and atypical outcome? Turing writes that both approaches are required and that much needs to be done, presumably in both computer education and in human education.

In other words, he closes with a call for pragmatism, a recognition and mobilization of many different forms of intelligence, and again he is warning us that imitation isn’t everything. Equating indistinguishable with intelligence may be a serious error. We read Turing as offering an implicit ethical admonition. Turing himself could not “pass” as “normal” in the world of his time and he wasn’t so concerned that machines couldn’t “pass” either. Intelligence will continue to pop up in surprising and indeed hard to recognize forms.

The history of the Turing test itself may illustrate how apparent cognitive deficits may in fact be entangled with underappreciated cognitive advantages. It is possible that Turing conceived of his imitation test precisely because he had so much difficulty “passing” and communicating himself. In social settings these facts were seen as disabilities but in the longer term they helped Turing produce this brilliant essay.
Ultimately, the conferring of status on another entity, machine, human, male, female, ethnic group, or pet is not an intellectual decision. It is an action based on the intentional stance we take toward the other. We confer a status on that entity. There can be rational justifications, but ultimately it is an action springing from deep within, generally hidden from our rational selves. We can be argued into changing our stance. But that doesn't mean our stance is a rational choice. It is a human choice, a choice coming out of us from levels which are inaccessible to our conscious mind. Pets become friends and family members. Some day in the future, intelligent machines will be raised in status to valued companions and, when they surpass us, as beloved guardians and guides.

During my lifetime I've seen great transformations in our relations to others. Those with mental disabilities have been more accepted. The barriers of race have come down. We have even come to accept that some animals have higher cognitive abilities that are dimly like our own. I fully expect this to continue and the barrier between humans and machines to blur.

In the past, viewing Down's Syndrome or Autistics as "mentally retarded" created a self-fulfilling prophecy as these kids were segregated and ghettoized and, sure enough, became so developmentally "challenged" that they were unintegratable into the broader society. That has changed. We now recognize that with strong intervention these kids can be challenged and respond with growth that, while it may not resulting in an "ideal" sociable human, can create somebody able to function in the broader society. We no longer warehouse them. We establish group homes or provide specialized services to allow them to have a degree of independence and a sense of a full life.

No comments: