“A Great science fiction detective story” – Ian Watson, author of The Universal Machine
NOW AVAILABLE for instant download! Click to find out more.
Days since the Centenary: 3
Days to the Bicentennial: 36,324
A few days ago I posted an entry entitled The Uncanny Valley and the ‘Flaw’ in the Turing Test, in which I disagreed with Terry Walby, the UK managing director at IPsoft, who wrote a post on the Wired Science blog entitled Why the Turing Test Is a Flawed Benchmark.
Walby argued that the Turing Test was flawed because machine intelligence ought not to be judged solely by comparison to human intelligence. He made the point that artificial intelligence could and should develop in ways that are unique to machines.
I agreed whole-heartedly with his point about machine intelligence, but took the position that the Turing Test was never intended to be an exhaustive test for any and every type of machine intelligence, built on the assumption that it must necessarily resemble human intelligence.
Rather, the Turing Test allows us to search for evidence of a specific type of machine intelligence that, once found, would be easy to recognize and particularly hard to refute or disqualify: the kind that resembles human intelligence.
Since I posted that piece we’ve had the long-awaited Turing Centenary, the June 23 celebration of Turing’s 100th birthday. Now that the hoopla’s died down and everyone has recovered from the hangovers they doubtless incurred playing Turing drinking games into the wee hours of June 24th, I want to return to that thought from a slightly different angle.
Consider another search for non-human intelligence: SETI is the Search for Extra-Terrestrial Intelligence. Its object is to locate, if possible, life on other planets.
This search has traditionally been conducted by scanning the sky for radio transmissions that would indicate the presence of alien intelligence on other planets. Stephen Hawking discusses the famous “wow” signal detected in the 1970s in the video below.
The search for a radio signal relies relatively little on preconceptions about the form of life we might find, having been constructed around the possibilities for telecommunication that are inherent in electromagnetic radiation and that are therefore available to a broad range of possible life forms.
But recently the search for life on other planets has found an additional focus: the planets themselves. This development has been made possible by technological innovations that make the detection of distant planets possible for the first time, and that increasingly allow us to make deductions about the environment on extra-solar planets.
In this context, we have begun looking for Earth-like planets, the ideal candidate having a rocky core that is the right distance from its local star to allow for liquid water and sufficient mass to permit it to hold an atmosphere. This type of search is constructed specifically to search for life that resembles us, because it looks for environments like the one in which we evolved.
The National Geographic video below gives some details about one candidate planet, Gliese 581.
But the search for life that resembles us through the detection of planets similar to ours doesn’t mean that we believe life must resemble us. It’s simply a technique for locating environments that could produce a type of life that we can readily recognize (because it’s like us) and that would leave little room for controversy over whether it constitutes life or not (again, due to its resemblance to Earth life forms).
There is no contradiction in carrying out a search of this kind while acknowledging that environments very different from Earth could allow for the evolution of life forms very different from those we’re familiar with, like those discussed in the video below.
The Turing Test is like the search for Earth-like planets. Out of the myriad of possible forms of intelligence, it is intentionally constrained to forms that are similar to human intelligence, not because that’s the only kind of intelligence that Turing believed could exist — he explicitly said otherwise — but because it’s the type that is easiest to recognize and most difficult to refute.
This isn’t a weakness of the Turing Test, it’s one of its great strengths.