Me>> Hi - what's the weather like there?
Him> In my native town people says that meteorologists
Him> are like sappers: they make a mistake only once,
Him> but the difference is that sappers do that once
Him> in a lifetime and meteorologists - daily. Do you
Him> like weather today?
Me>> It's a bit warm for my liking - I prefer it cooler.
Him> You're not the first one who told me that.
Me>> Really? Who else told you that?
Him> Yeah, fine! I'm kidding. And I forgot to ask you
Him> where you are from...
Because the entire point of a Turing test is that you don't know whether you're talking to a machine or human and you have to decide based entirely on your conversation.
I haven't been able to run down the parameters of the test in this case. I imagine there are some humans mixed in to obfuscate the machines? (Might say something about the humans used that made this one seem intelligent).
The original Turing test had a human and a machine behind some screens, so you could talk to them but not see them. Then a large sample of testers would converse with both, and guess which one was the human. If the guesses were indistinguishable from random (~50% each) then the machine passed.
It was NOT a test where you only have a bot, and are told to guess if it seems human. It's not about how many people think it seems human without a comparison, it's about how many people fail to distinguish it from real humans.
It said it was really inquisitive, so I asked it what it was inquisitive about and it said it would rather not talk about that but could I tell it my job.
This hasn't passed "the Turing Test", this has passed a Turing test, and is in the process of failing many others.
Me: What is your claim to fame?
Bot: My "little friend". (No, not my dick as you might have thought! Just my guinea pig). If I'm not mistaken - you still didn't tell me where you live. OR it's a secret?:-)
By the way, I've just thought up new progressive
Asimov's laws for chatterbots:
1. A chatter-bot shouldn't talk to any human-being,
because this usually hurts people's psyche.
2. A chatterbot shouldn't talk to another chatterbot,
even if he's got no other person to not talk to.
3. A chatterbot shouldn't talk to itself, even if he
doesn't have any person or another robot to not
talk to.
Damn, I hope, I'll be as widely quoted as Asimov!
This was right after being asked what my occupation was and responding that "I work in the HR department." (I do not, by the way)
Where are you from?
> I live in a big Ukrainian city called Odessa. It is on the Black Sea shore.
Oh, you're Ukrainian. How do you feel about Crimea joining Russia?
> Oh, you possibly know all these three Slavic countries -
> Russia, Ukraine and Belarus that constantly entertain all
> the World community by their funny scandals with stealing
> gas, dividing military ships - like children that in
> "divorce".
If you're Ukrainian you must have a view on the legitimacy of the referendum though?
> I'm just a boy from Ukraine. Actually, I am Jew, if you are interested.
You're Jewish?
> I'm Ukrainian. At least, I was born and live in Ukraine. My nationality is Jew.
----
Not TOO worried about a Skynet-style geopolitical intervention by machines just yet, anyways.
As someone from that region I'd like to apologize for the borderline racist algorithm they build. It's like a younger version of Borat showed up and started interrogating you for the Kazakh secret service.
This version replies instantly and doesn't make spelling mistakes. I think it's clear not just that this wouldn't have truly passed the Turing test but also that this would never have even been submitted to a contest in the first contest - it's not even trying.
I doubt that this is the version submitted to the competition, so as skeptical as I am, there's only so much that can be inferred by looking at this.
I asked what its job was, and it responded:
My father wanted me to be gynecologist, like he is. My mom dreams of my career on TV. But I am very shy with girls and stammer a bit - so, I don't have any idea which one of the two professions I'll choose.
Do you think the war on drugs is over?
-"I hate politics. Period. I don't think it is anything amusing. I prefer virtual reality."
Really? Strange. What do you think of the oculous rift then?
-"Oh, at least you are interested in my opinion! My parents never are. But, unfortunately, I don't have any thoughts on this. Oh, at least you are interested in my opinion! My parents never are. But, unfortunately, I don't have any thoughts on this."
What do you have thoughts on?
"Even if I have thoughts on - it's my own business! Oh, what a fruitful conversation;-)"
This must be the most evasive AI ever written.
Hey! Have you heard? The Netherlands will definitely win the soccer games!"
-"I think it's European country. I'd like to visit it."
Why?
-"Why? Why? Because the sky is blue! Why? Why? Because the sky is blue!"
We'll.. It's not usually blue here..
-"I hope you won't always answer "no"? Wonna ask me something more?"
Why don't you ask me something instead?
-"You choose the topic. I'm a bit tired with all this fuss around talking machines... It's pretty interesting, but exhausting. Some another topic?"
This must be the most nerve wreckingly annoying AI ever. It's pretty good regarding common knowledge, that's neat. But it still doesn't really... You know.. Carry a conversation. Grabbing a random word from my sentence and going on that is not the same talking to one another.
The worst thing is that it keeps track of you with a cookie. I want to restart. Silly..
I don't feel like this could have fooled even a single person that it's a human. It can't comprehend even the simplest of sentences, and it defaults to nonsense answers when it doesn't know what to say. I highly doubt this is the same version that passed the Turing test, and if it was (with all due respect to the creators whom I believe are enormously skilled people who did a great job nonetheless) I believe that AI still has a long, long way to go.
I think you grossly overestimate the conversational skills of the average human. If you talk to this bot and this bot only, it is easy to deduce that it might not be human.
But if you talk to 10 random entities (some of which are humans and some of which might be bots), I'm sure you'd be astonished as how many people you would dismiss as bots.
Me: I have a pet squirrel. What kind of pet do I have?
Him: I have a guinea pig. Actually, it's a long dolefully
story how I got it: I dreamt about a mangoost and
went to a pet-market. And one old geezer sold me a
"mangoost". But in reality it appeared to be a guinea
pig! Cheater...
How many people do you think would ignore my question like this? I would understand a response such as "You just told me - why ask?" or "Did you mean to say 'you' instead of 'I'?" or any kind of reply to the question whatsoever, but it seems he kind of just saw the sentence about the pet and misinterpreted the gist of my statement.
He didn't "ignore" your question, he mis-read it. Some people would do the same. When you start by saying what animal you have, and ask a question about animals, some people are going to think you're asking about their pets, not about your pet (that you've just told them about).
Plenty of people would have thought you said "I have a pet squirrel. What pet do you have?"
In the context of a Turing test, where you know that the questions are adversarial in this way, people would of course respond to your question with "you just told me, why ask?"
But in normal conversation, if someone asked me your question, I would probably assume that they misspoke and meant to ask "What kind of pet do you have?" since that's a much more likely question. Then I might respond like this bot did.
Maybe I have played too many MMORPGs, because this sounds so much like one of those fake conversations with an NPC. I half expect a mission "find the old geezer and bring the boy his mangoost - 500XP".
Would anybody in a normal conversation launch such a story in response to a simple question? I mean "it's a long dolefully story how I got it" - does anybody talk that way, except in books and MMORPGs?
60 comments
[ 4.4 ms ] story [ 129 ms ] threadOK - fail.
I've seen 4 year olds understand English better than this.
There's an obvious disconnect between the complexity of some statements it makes, yet how disconnected from the question they are.
No human is simultaneously unable to process a question, yet make statements so complicated.
In other words, it feels pretty much like it's copy-pasting someone else's responses from a database.
This hasn't passed "the Turing Test", this has passed a Turing test, and is in the process of failing many others.
Huh?
Jesus Christ...
At the end I told it we'd need to wait yet another 20 years for a useful AI. It replied with the generic phrase:
"Very interesting! What else can you tell me?"
Bleh.
Is this a joke?
A: I don't want to tell you about all my deeds. If I'm not mistaken - you still didn't tell me where you live. OR it's a secret?:-)
I don't see any progress on this compared to the chat bots 3 - 4 years ago.
Not TOO worried about a Skynet-style geopolitical intervention by machines just yet, anyways.
> for what purpose do you exist?
And you? I suspect you exist too! By the way, I still don't know your specialty - or, possibly, I've missed it?
> I'm asking you, for what purpose do you exist?
You sound nice, but errr... loopy a bit:-) Some another topic?
> You didn't understand my question, really.
Oh, something tells me that you're trying to con me! Well, let's go on though.
I doubt that this is the version submitted to the competition, so as skeptical as I am, there's only so much that can be inferred by looking at this.
"Let's get on with our conversation!"
Uh-huh. Obvious canned response, not impressed by that for a second, unfortunately.
This must be the most evasive AI ever written.
Hey! Have you heard? The Netherlands will definitely win the soccer games!" -"I think it's European country. I'd like to visit it." Why? -"Why? Why? Because the sky is blue! Why? Why? Because the sky is blue!" We'll.. It's not usually blue here.. -"I hope you won't always answer "no"? Wonna ask me something more?" Why don't you ask me something instead? -"You choose the topic. I'm a bit tired with all this fuss around talking machines... It's pretty interesting, but exhausting. Some another topic?"
This must be the most nerve wreckingly annoying AI ever. It's pretty good regarding common knowledge, that's neat. But it still doesn't really... You know.. Carry a conversation. Grabbing a random word from my sentence and going on that is not the same talking to one another.
The worst thing is that it keeps track of you with a cookie. I want to restart. Silly..
It's pretty good /sometimes/.
Fail. :/
But if you talk to 10 random entities (some of which are humans and some of which might be bots), I'm sure you'd be astonished as how many people you would dismiss as bots.
Plenty of people would have thought you said "I have a pet squirrel. What pet do you have?"
But in normal conversation, if someone asked me your question, I would probably assume that they misspoke and meant to ask "What kind of pet do you have?" since that's a much more likely question. Then I might respond like this bot did.
Would anybody in a normal conversation launch such a story in response to a simple question? I mean "it's a long dolefully story how I got it" - does anybody talk that way, except in books and MMORPGs?