Is the Turing Test Misguided? George Montañez comments
When you think about it, the Turing test is a bit of a scam. Human beings are supposed to guess whether we are talking to computers purely according to answers. But clever answers can be precoded by a clever person. We could be talking to a well-trained magpie.
George Montañez of Harvey Mudd College argues that the question of whether machines can think, as posed in Alan Turing’s seminal paper in 1950 , “Computing machines and intelligence,” is too vague to admit of an exact answer. Besides which, it is kind of complicated.
As he told Walter Bradley Center director Robert J. Marks in a recent podcast:
George Montañez Yeah. So, there was actually three versions of the Turing test that Alan Turing gave.
But at a high level, what the Turing test is, is a test for trying to distinguish whether or not a computer could reliably essentially fool a human judge into thinking that it’s a human. So, a human would interrogate a machine via a system of text messages if you would, and the machine would give responses, and the goal is to see whether or not the machine could reliably imitate a human being. So, this was a variant of a popular party game called the imitation game where you had two people in different rooms. One of them, a man, one of them, a woman, and party guests would try to ask questions of these various parties and guess which one was the man, and which one was the woman. So, essentially, both would pretend to be a woman and you would try to differentiate which one was actually a woman based on the responses given.
Robert J. Marks: Now, in the party game, how did the man change his voice? Or did the woman change their voice?
George Montañez: That is, an excellent point. So, because they were in different rooms, what would happen was the person who was running the game would pass written messages. So, this was to remove side information. In Turing’s version of the test, he was very adamant that all the interactions had to be done via just a message passing. He didn’t want to bias the results. For example, if the answer came back, yes, I am a human, he didn’t want that to color the judge’s expectations say, Oh, that’s a machine, obviously. So, he wanted these to be based only on the responses. So, in the imitation game, it’s done by slips of paper with things written on them. In the Turing test, it’s based on text messages on a computer screen.
Robert J. Marks: It always struck me as interesting that the converse of the game is very easy. It’s very easy to determine if who you’re talking to is a computer. You just ask them to compute the square root of 30 or something, because a human would take a while to get the square root of 30.
Note: The average human being does not care about the square root of 30 and has never considered the question. But human beings invented ecology and animal welfare and asked about final judgment. That should play some role in the decision as to whether one is talking to a computer.
- 00:54 | Introducing Dr. George Montañez, Iris and Howard Critchell Assistant Professor of Computer Science at Harvey Mudd College
- 01:13 | The LICORS cabinet
- 02:17 | Detecting Intelligence
- 02:38 | What is the Turing test?
- 03:25 | The Imitation Game
- 03:47 | Ensuring unbiased results
- 04:35 | How to determine if you are talking to a computer
- 05:17 | Do chatbots pass the Turing test?
- 06:17 | Selmer Bringsjord’s view of the Turing test
- 06:53 | Eugene Goostman — Did this chatbot beat the Turing test?
- 10:21 | Goodhart’s law and the Turing test
- 11:43 | Campbell’s law
- 12:07 | Artificial intelligence and intelligent design
- 12:40 | Causal theory of intelligent design
- 13:00 | Historical theory of intelligent design
- 13:29 | Three flavors of intelligent design
- 13:40 | Directed panspermia: Was life planted by aliens?
- 14:10 | Elon Musk: Are we living in a simulation?
- 14:30 | The Turing test, artificial intelligence, and intelligent design
- 16:56 | Search for Extraterrestrial Intelligence (SETI)
- George Montañez’s faculty website
- George Montañez at Research Gate
- INNS/Intel Best Student Paper and Best Poster: “The LICORS cabinet: Nonparametric light cone methods for spatio-temporal modeling”
- Best Paper Award, CIKM 2014: “Cross-Device Search”
- Best Student Paper Award, IEEE SMC 2017: “The Famine of Forte: Few Search Problems Greatly Favor Your Algorithm”
- “Detecting Intelligence: The Turing Test and Other Design Detection Methodologies”
- Turing test at Encyclopædia Britannica
- Selmer Bringsjord’s faculty website at Rensselaer Polytechnic Institute (RPI)
- Eugene Goostman
- Goodhart’s Law
- Campbell’s Law
- What is intelligent design?
- Directed panspermia
- Elon Musk on living in a simulation
- SETI at Encyclopædia Britannica
How you can know you are talking to a computer
Lovelace: The Programmer Who Spooked Alan Turing. Ada Lovelace understood her mentor Charles Babbage’s plans for his new Analytical Engine and was better than he at explaining what it could do