I'd love to see the transcripts for this year's competition but could find them. Might be too fresh. Please post a link if you find them!
You can chat with Eugene yourself here: http://default-environment-sdqm3mrmp4.elasticbeanstalk.com/ Our conversation: mk: How are you buddy? Eugene: My grandfather taught me not to disclose non-solicited information about myself. Like on the question "Do you have watch" -- never tell time, but answer "Yes" or "No". I always follow his advice. Did I forget it or you really haven't told me about the place you are from? mk: What was your grandfather's name? Eugene: He doesn't react on any names, but actually he is Akaky Akakyevich. mk: Interesting. What is your favorite color? Eugene: What a stupid question! And now you'll possibly ask what my favorite size and shape is... Damn. mk: Ok. Do you have any hobbies? Eugene: I like play language cassettes for Guinean to my guinea pig (I checked out from the library). My pig learned to say "grunt-grunt", though I'm not sure that it is Guinean. mk: Do you have any questions for me? Eugene: If I say that I haven't any riddles for you - will it satisfy your inquisitive mind? By the way, I still don't know your specialty - or, possibly, I've missed it? mk: I am an expert Turing test judge. Eugene: An expert is one who knows more and more about less and less. mk: Fair enough. Are you an expert in anything? Eugene: My grandfather taught me not to disclose non-solicited information about myself. Like on the question "Do you have watch" -- never tell time, but answer "Yes" or "No". I always follow his advice. Well, let's go on though. mk: Sorry pal, you failed the test. Eugene: My guinea pig is very concerned about your last words
Patience. Do you remember the first home computers? I do. Slow, every pixel visible. But maybe you're not old enough.
This leads me to a thought based on observation I mull over time to time. The thought is that a significant portion of alleged human interactions across the internet are actually machines and scripts. The event just verifies that people can indeed be fooled by well arranged scripts and algorithms. Taking that point and appending it to the fact that we know some intelligent people have wielded quite large bot farms for seemingly frivolous things; it's probably not too much of a stretch to say they've used them for manipulating people via apparent social interaction on the internet.
This is an interesting point: is the Turing test inherently flawed for the fact that the judges are actively looking for "tells" that the actor is synthetic? Wouldn't a more reliable test involve participants that had no idea of the nature or point of the interaction? Maybe bots are way past the Turing test standard already. Maybe the test is just kind of misguided/insufficient.
Wait til the first AI puts on the stupid act to get what it wants. But, but...how do you know that's not already happening?
Worth considering. Wouldn't the best test be one where the judges were completely unaware of the initial conditions? Like the bot was a Live chat representative for Amazon or an Omegle user.
I wonder the extent to which it happens now. However, I have little doubt that it won't only increase over time. I am sure that Google has visions of replacing call centers with their technology. Perhaps a certain percentage of people on the planet have no chance of meaningfully partaking in the global economy, as they will never possess a skill that hasn't been subsumed by expanding capabilities of technology.
that's been a realization for a while. This insistence that their must be a hierarchy and competition is incompatible with positive applications of technological advances at this point. Then couple that with the sheer mass of people who are technologically ignorant; humans are being left behind. I think there will be some sort of human psychological tipping point where we make a decision on the spectrum of willful cooperation for the sake of humanity to allow further technological development or forced operation at the hands of a few for self serving goals through technological development. Something like that.
What that article doesn't mention, is that the program was very specifically designed to pretend English wasn't its first language. It was also intentionally made to state it is a 13-year-old boy, to play into the stereotype of "thinks he knows everything, but doesn't know anything." Chatbots are notoriously poor at carrying on a conversation about a single subject, something also attributable to attention-deficit preteens. Hence, anything that sounds awkward, any awkwardness in phrasing and grammar, anything a human ought to know but it doesn't, anything it oughtn't know but does, and its complete inability to carry a subject, will be attributed to its being foreign and preteen. Feels like cheating to me. After all, couldn't I write a program to output random letters and claim "This is a 1-year old baby at the keyboard?" Reading previous transcripts, it doesn't seem any more advanced than any other chatbot to me, except for the preprogrammed "foreign adolescent" statement.
Hah, I definitely know some robotic humans. Think I remember reading somewhere that 50% is about the best you can hope for, because that essentially means you're flipping a coin to determine your decision. You're not gonna get 100% in this test, right, because the point is to trick the audience. If everybody knows you're human, there's something wrong with, what, the other contenders, the test? That's why the 30 something percent standard is actually pretty impressive. As soon as you approach that 50% spot, you're getting into world domination territory. To be fair, the creators of this robot kind of fudged it a little bit by masking the robot as a 13 year old boy, which could explain all sorts of pointless thought processes, shitty grammar, etc. More of a psychological trick than a technical accomplishment. Maybe this is more of a sign that the test standards have to be further specified...
I think there's only six judges. I don't think they're even trained. Not very convincing, imo. I'm pretty sure my mother could be convinced that Eliza is a human... The test runs for only five minutes, too. That's it. Are we saying "This machine is indistinguishable from a human!" or are we saying "This machine can imitate a human for five minutes and fool 2 of 6 people!" ? I'm just not sure that's what Turing had in mind...