TOTW: The Basic Threshold Into Sentient AI Has Been Passed – Kinda
For the last 60 years, artificial intelligence experts and computer experts have been striving to solve one puzzle. Or really, pass one test. The founder of the modern computer, Alan Turing (the guy who cracked the Enigma code) created a test way back in 1950 as the threshold onto efficient artificial intelligence. The test is fairly simple. All the computer had to do was convince a panel of judges that it is a real person. Any type of person. A 60 year old professor. A 20 year old computer scientist. Doesn’t matter. And the computer doesn’t even have to fool all of the judges. Just 1 one third of the panel. Even though this seems like this would be fairly easy, it hasn’t been done in 60 years.
This test was Turing was to test the age old question: can computers think. When this test is passed, he though we would be somewhere near that age. There have been some sketchy wins in the past, not really confirmed or in an official setting. But, in the Turing Test 2014 official event, a program called Eugene Goostman completed the test successfully, tricking 33% of the judges into thinking he is the average 13 year old Ukrainian boy. The description given by his “parents” about Eugene is that “His Dad is a gynecologist and he has a hamster”. Seriously. That’s the description they gave him.
Now, before you go around screaming about robotic armageddon in gas stations I should remind you that this means a pretty much nothing. First of all: he is not a sentient being. He isn’t even a being. Eugene is a chatterbot. A program that responds to you in chat, posing as the person they are testing. A chatterbot would have a hard time taking over the world. Secondly, he only convinced 30% of the judges. As the Guardian wrote, 30% of Americans think the Bible is literally true. That doesn’t make it so. It is an achievement, being the first in history to complete this task made by the founding father of technology, but it’s not game changing. When 100% of the judges are fooled, then watch out.
Plus, the makers of Eugene were pretty tricky about how they won. Because the target was a 13 year old, they made it acceptable that Eugene could just basically reject the question by saying he didn’t know. Of course, a normal 13 year old would probably do that, and making a program admit to not knowing everything is very smart in trying to duplicate a person. Oh, and english is his second language, so his answers don’t have to perfectly intelligible. Yet, despite all these facts disregarding Eugene’s accomplishment, it still is a fairly big event for all AI geeks out there. After all, this test was made by Alan Turing, computer genius extraordinaire. It may not mean we have sentient computer beings from sci-fi movies, but it’s not nothing. A man-made object fooled a man. Sheesh, we humans are dumb.
If you want to see if Eugene can fool you, here’s your chance. Eugene’s creators put Eugene’s code online, so anyone could try it. I spent a couple of minutes with Eugene, and I wasn’t fooled. I asked him is he was sure if he was 13, and he answered yes then started pressuring me to tell him where I lived. I told him I had a giant candy cane and he said candy is a geek. Obviously, there’s a way to go until computers will be functional as a replacement human.