Yesterday, a computer program successfully passed the Turing test for the first time. As is not uncommon for breakthroughs, it was a controversial victory, with plenty of detractors pointing out the various reasons that this wasn’t that big a deal after all. What is one to think! Let’s talk it out.
What’s everyone so worked up about?
This past weekend, a computer program successfully convinced 33 per cent of a panel of judges that it was a 13-year-old boy named Eugene. That makes it the first time a machine has passed the Turing test, which has been around since famed computer pioneer Alan Turing introduced the concept in 1950. Big milestone!
OK, but what’s the Turing Test?
Good question. There seems to be some confusion over what exactly the Turing Test is meant to measure. Its purpose isn’t to indicate whether a machine has achieved artificial intelligence or sentience; this ain’t Skynet we’re talking about. In the most simplified terms possible, the Turing test is meant to measure whether a machine can successfully imitate a human. Think clever chatbot more than I, Robot. It’s also a pretty low bar to pass; just 30 per cent of testers need to be fooled.
Why is this win so controversial?
One of the biggest caveats is that the Turing test itself isn’t so much a test of computer intelligence as it is human gullibility. A bad chatbot might luck its way to victory if the judges aren’t familiar with tell-tale signs of chatbot-ness. That’s usually of less importance when your panel includes experts in the field of computer science. In this case, it included an actor from Red Dwarf and a member of the House of Lords, both of whom are incredibly accomplished and by all indications brilliant minds, but not specifically trained in this field.
On top of that, while Eugene technically passed the Turing test — no argument there — the way it did so was pretty sneaky. Its creators decided that rather than try to mimic, say, an Oxford professor of Classics or even just a normal human being, they would seek out the lowest possible form of recognisably human conversation: A teenage boy. Not only that, but a teenage boy for whom English is a second language.
In other words, Eugene passed the Turing test by imitating a human being whose answers would be barely cogent under the best of circumstances. It’s not quite Kirk and the Kobayashi Maru, but it’s an impressively devious bit of bar-lowering.
So… we shouldn’t care?
Here’s the thing: Even with all of these caveats and asterisks, this is still an absolutely stunning breakthrough. Eugene is not the first chatbot to try to sneak its way to victory, but it’s the only one to have actually done it under circumstances that are respectable. And while the panel may not have been full of Nobel Laureates, most of the humans a chatbot would have to fool aren’t geniuses or specialists either. This is a huge accomplishment. Your enthusiasm is warranted.
If you must experience disappointment of some sort, direct it at the Turing test itself. It’s a useful rule of thumb, but it’s subjective by its very nature, and sets a pretty low bar. Being mad about Eugene’s victory is like being mad that field goal posts are too wide when someone makes an extra point. Don’t hate the chatbot, hate the chatbot game.
Don’t worry, HAL 9000’s not about to show up at your doorstep. But this signals that we’ve officially entered an era of computing where programs are able to trick you into thinking they’re human — even if those humans are 13-year-old boys with limited vocabulary.
The implications extend as far as the imagination will let them. Troubleshooting chatbots, marketing chatbots, weird sex chatbots, normal sex chatbots, troubleshooting chatbots that market weird sex. And maybe even more importantly, we can finally move on from Turing and set our sights on the next great computing milestone. That’s when we’ll finally have a Skynet worth cowering to.