The Origin of Artificial Intelligence, AKA: Is This Thing Alive?

Alex Johnson - Content Writer

Tags:

Ah, the age-old question: “Is this thing alive?”

When I was a kid, I’d spend my summer days pulling little Lovecraftian horrors out of the sea and asking my parents that very same question. Almost always, the answer was: “Yes, put it back.”

British mathematician and logician Alan Turing asked a similar question. Only, instead of asking it about spider crabs and devil’s purses, he asked it about computers. Also, he didn’t ask his parents, because he was a grown-ass man. Instead, he devised a test to answer this question. Or, more specifically this question: “Is this thing intelligent?”

It’s an important distinction. The things I dragged ashore in the 1980s were indeed alive (or had been), but they didn’t possess intelligence, not as we humans understand it. And Turing wasn’t really interested in the intelligence (or lack thereof) of spider crabs — he was a logician, so his interest was in intelligent behavior, specifically in machines. If a machine could exhibit such behavior, then who’s to say it isn’t alive? Not alive in the biological sense, but in a philosophical or ethical sense. From the human perspective, intelligence speaks to a higher form of life, and Turing believed computers would reach this stage in the near future.

Who Was Alan Turing?

Before we tuck into Turing’s work, let’s learn a little bit about the man.

Alan Turing was born in 1912 in London, England. Through private schooling and his mathematics studies at the University of Cambridge (ever heard of it?), Turing got himself nice and smart, and graduated in 1934.

Alan Turing, a human . . . or was he? (He was.)

After that, Turing jumped right into it, publishing his seminal paper “On Computable Numbers, with an Application to the Entscheidungsproblem” in 1936. All you need to know about the word Entscheidungsproblem is that A) it’s German for “decision problem,” and B) if you really want to understand it, ask a mathematician—I studied poetry in college.

The gist, however, is that there were “computers” doing math in the 1930s, but the word “computer” meant an actual person crunching numbers using rote methods. You know, like a real dream job.

Since mathematical clerks were still the only way to solve the big mathematical research problems, the purpose of the Entscheidungsproblem was to find an effective method for “determining exactly which mathematical statements are provable within a given formal mathematical system and which are not.”1

Basically, Turing was looking for a decision method that would make human-based computing more effective. For our purposes here, the important thing is this: It was while working on the Entscheidungsproblem that Turing invented his universal Turing machine—a hypothetical, abstract computing machine that presaged the logical principles of the digital computer.

Turing’s argument, now called the Church-Turing thesis (another mathematical logician, Alonzo Church, was also working on this stuff), was that his Turing machine could compute anything a human could. This was an important claim: the Turing machine had computational limits, and if it could compute anything a human could compute, that meant humans had computational limits, too. And to support this assertion, I present Exhibit A: I can’t really do long division.

So, you can see how Turing might have been interested in the relative limitations of both human and non-human computational methods.

Turing began studying for his PhD in mathematical logic at Princeton University. He wrapped that up in 1938, returned to England, and promptly joined the government’s cypher school. England went to war with Germany in 1939, and Turing spent World War II breaking the schieße out of German codes. He worked with the team that created the “Bombe,” the code-breaking machine that decrypted messages sent by the Germans and encoded with their Enigma cipher machine. Their work allowed the Allies to decode around 39,000 intercepted messages each month, which made it a real tide turner in the war effort.1

The First Computer(s)

With the whole Hitler thing handled by 1945, it was time for Turing to get back to replacing us all with machines. The National Physical Laboratory (NPL) in London snatched Turing up and put him to work creating an electronic computer.

Turing’s computer design was called the Automatic Computing Engine, or ACE. His original design called for a superior machine that would have had way more power than any of the other early computers would, plus it would come pre-loaded with Minesweeper. Unfortunately, his coworkers were less optimistic, so the end result was a much more modest computer: the Pilot Model ACE, built in 1950. No Minesweeper.

I know it’s not what Turing wanted, but at least it was more intuitive than Windows 8. I still don’t get why my computer in 2012 had a smartphone interface.

 

Even with their technology bar lowered, the NPL lost the race: the Royal Society Computing Machine Laboratory at the University of Manchester built the first working electronic stored-program digital computer in June of 1948.

Turing was bummed, so he jumped ship and joined the winning team as deputy director in 1948. It didn’t hurt that his new team were already fans of Turing’s pre-war work—his Turing machine concepts were a big influence on the Computing Machine Laboratory’s computer.1

Artificial Intelligence and ‘The Imitation Game’

Turing was an early supporter of the hypothesis that the human brain is, in part, a digital computer. His idea was that we’re born with an “unorganised [sic] machine” that becomes organized “into a universal machine or something like it” through training.

As this hypothesis blurred the line between a human computer and a mechanical or digital one, Turing created his Turing test.1

The Turing test—or “The Imitation Game,” as he proposed it—was meant to ultimately answer Turing’s questions about artificial intelligence.

The game had several iterations, each with various contestants.

Man vs. Woman

This version involved no computer intelligence at all. Rather, it featured two contestants—a man and a woman—as well as a human judge. Each person interacted with the others remotely through computer terminals, to avoid any physical interaction. Communicating through the computers, the judge needed to determine which of the two contestants was the man. The male contestant’s goal was to convince the judge that he was the man, while the woman’s goal was to trick the judge into thinking she was the man.

Fortunately for us humans, no computer will ever pass our CAPTCHA tests.
It’s our best line of defense.

I know what you’re thinking: “Why do all the contestants in this game want to be the man so bad?” Well, Turing proposed The Imitation Game in 1951, a time when men were men, and women pretended to be men via computer terminals, I guess.

At any rate, this version of the test was merely a stepping stone to the variants that followed.2

Human vs. Machine

This was the more famous version of the Turing test. This time, the contestants included a human of any gender and a computer. Again, a judge was tasked with figuring out which was which. The human contestant would play a role similar to the man in the first version of the game, attempting to honestly convince the judge that they were the human. The computer would fill the role previously assigned to the woman, and would try to convince the judge that, no, they were the human.

Turing’s proposal was this: if the judge had a success rate of under 50 percent for correctly picking the human contestant, then that meant the judge would be as likely to pick the human as they were to pick the computer. If that was the case, then it would mean the computer being tested was a passable simulation of a human being—at least in terms of communication without physical clues—and therefore was intelligent.3

Single Contestant

More recently, the test was again modified. This time, there was only one contestant and a judge. The contestant might be a human or a computer, and the judge’s job was to guess which it was. I play a simpler version of this game every time I get a new Twitter follower—are they a person, or just a bot? While I can never be 100 percent sure, they’re definitely always bots.4

Beating the Game

Turing’s prediction was that, by the year 2000, computers would be able to play his Imitation Game so well that the average judge would have less than a 70 percent chance of making the right decision after five minutes of questioning.

Some computer programs have come close to stumping enough judges to meet this prediction, but they’ve done it by playing the game smart, rather than hard.

For example, “ELIZA,” written in 1966, did pretty well by simulating a Rogerian psychotherapist—the type of therapist who takes a passive conversational role, asking leading questions and doing very little talking. By adopting this character, ELIZA was able to avoid a lot of heavy lifting. The fact that ELIZA didn’t talk much and asked leading questions made sense to some judges: that’s what Rogerian psychotherapists do.5

A more recent near-success story happened in 2014. The computer program was known as “Eugene Goostman.” Eugene was a fake 13-year-old kid from the Ukraine. The genius behind this program was that its backstory explained its communication flaws.

Eugene Goostman. I dunno, looks like a real kid to me.

So, when a judge asked Eugene a question and received a batsh*t non sequitur and a smiley emoji instead of a direct answer, well, that just made sense. The kid was 13 and wasn’t so great at English. And that’s how Eugene fooled 30 percent of its human judges after five minutes of questioning.

But this didn’t mean Eugene Goostman passed the Turing test. To do that, Eugene would have needed to fool half of the judges. Rather, Eugene, through some trickery (which, to be fair, was the point), merely convinced enough judges to meet Turing’s prediction for computers in the year 2000, only 14 years late.6

Cool, But What Does It All Mean?

The real question for us, if not for Turing, is this: did Eugene, or its predecessors like ELIZA, actually exhibit intelligence? For all the logical empiricism behind the Turing test, it measures how easily humans are fooled as much as it measures computer intelligence; the deeper questions are more subjective.

Humans aren’t perfectly logical, nor do we use language perfectly, as my editor could tell you. Consider the idea that an artificial intelligence that’s too smart might not pass for a human, either. Clearly exceeding human capability could be a quick way for a brainy artificial intelligence to flunk the Turing test.

Yeah, we’re definitely gonna pay for this sh*t.

The Imitation Game, or Turing test, is ultimately designed to determine whether or not a computer program can be indistinguishable from a human contestant on a relatively consistent basis. So far, computer programs have arguably failed because they aren’t capable of reaching a human level of sentience, only able to occasionally compensate by masking their flaws with programmed human characteristics.

But I have a dream that, someday, artificial intelligences will fail the Turing test because they’re clearly way more sentient than any human could be. Of course, by that point, we probably won’t be allowed to run tests at all. Instead, our artificial benefactors will keep us safe in bio-pods while they devise fun little games that they can force us to play.

I give it another 50 years. If I’m wrong, well, Turing got his prediction wrong, too, so I’m in good company.

Share this article... your friends will love it too ❤️

Notes & Gossip 📌

  1. Copeland, B. J. (Last updated February 14, 2018). Alan Turing: British Mathematician and Logician. Retrieved from https://www.britannica.com/biography/Alan-Turing
  2. Reingold, Eyal, and Johnathan Nightingale. (Accessed March 29, 2018). The Turing Test. Retrieved from http://www.psych.utoronto.ca/users/reingold/courses/ai/turing.html  
  3. Reingold, Eyal, and Johnathan Nightingale. (Accessed March 29, 2018). The Turing Test. Retrieved from http://www.psych.utoronto.ca/users/reingold/courses/ai/turing.html  
  4. Reingold, Eyal, and Johnathan Nightingale. (Accessed March 29, 2018). The Turing Test. Retrieved from http://www.psych.utoronto.ca/users/reingold/courses/ai/turing.html  
  5. Reingold, Eyal, and Johnathan Nightingale. (Accessed March 29, 2018). The Turing Test. Retrieved from http://www.psych.utoronto.ca/users/reingold/courses/ai/turing.html  
  6. Lopatto, Elizabeth. (2014, June 10). The AI That Wasn’t Why ‘Eugene Goostman’ Didn’t Pass the Turing Test. Retrieved from https://www.thedailybeast.com/the-ai-that-wasnt-why-eugene-goostman-didnt-pass-the-turing-test

Scholarly Shout-outs 🌟

  • CNET. (2015, March 31). What is the Turing Test? Retrieved from https://www.youtube.com/watch?v=sXx-PpEBR7k
  • Copeland, B. J. (Last updated February 14, 2018). Alan Turing: British Mathematician and Logician. Retrieved from https://www.britannica.com/biography/Alan-Turing
  • Encyclopaedia Britannica. (Accessed March 30, 2018). Turing test: Artificial intelligence. Retrieved from https://www.britannica.com/technology/Turing-test
  • Lopatto, Elizabeth. (2014, June 10). The AI That Wasn’t Why ‘Eugene Goostman’ Didn’t Pass the Turing Test. Retrieved from https://www.thedailybeast.com/the-ai-that-wasnt-why-eugene-goostman-didnt-pass-the-turing-test
  • Reingold, Eyal, and Johnathan Nightingale. (Accessed March 29, 2018). The Turing Test. Retrieved from http://www.psych.utoronto.ca/users/reingold/courses/ai/turing.html
  • TED-Ed. (2016, April 25). The Turing test: Can a computer pass for a human? – Alex Gendler. Retrieved from https://www.youtube.com/watch?v=3wLqsRLvV-c