The Digg Crew wants to hear your thoughts!
Please take our short survey about Digg and potential feature ideas.
Almost human: Interview with a chatbot
technology.newscientist.com — Every year the Loebner Prize for artificial intelligence is awarded to the chatbot software able to converse most like a human.
Michaelabehsera- 555 diggs
- digg it
- PoliticalWinds, on 10/14/2008, -0/+25Sometimes I wonder how many comments posted on Digg are done by bots.....
- Roryking, on 10/14/2008, -0/+32How do you respond when people pose this question to you?
- Shootfast, on 10/14/2008, -2/+160110001001101111011011110110001001110011
- jamesdew, on 10/14/2008, -1/+11I think "How do you respond when people pose this question to you?" could be a new meme
- Neo829, on 10/14/2008, -0/+2@shootfast:
I like the way you think.
- sc0rpi0n, on 10/14/2008, -0/+2Well, I'm a robot with a built-in chat module That's better than long explanations any day.
- 5xSTUN, on 10/14/2008, -0/+9Tell me more about why you wonder that.
- JKAL, on 10/14/2008, -0/+3To answer your question, none. All comments are done by real people and not bots.
To answer your question, none. All comments are done by real people and not bots.
To answer your question, none. All comments are done by real people and not bots.
To answer your question, none. All comments are done by real people and not bots.
To answer your question, none. All comments are done by real people and not bots. - Suspected, on 10/14/2008, -0/+1 I can't really tell you much about how interactive assistant tick. At the moment I'm in a group therapy class trying to get to the bottom of that.
- Roryking, on 10/14/2008, -0/+32How do you respond when people pose this question to you?
- wolfeater2, on 10/14/2008, -2/+8that makes pedo bear so happy.
- Mujokan, on 10/14/2008, -1/+80I can't believe Elbot fooled anyone, quite frankly...
- sc0rpi0n, on 10/14/2008, -0/+2I think your justification hits right at the crux of robotic and humanic relations.
- Mujokan, on 10/14/2008, -0/+4Yes, I'd quite like to hit a robot in the crux.
- Aslan72, on 10/14/2008, -0/+2That's a startling proposition for flesh to metal.
- Arasaka, on 10/14/2008, -0/+12All the "wit" of chatbots are really just the creators finding new, slightly varied prescripted ways of making the bot answer "NO U" to every damn question.
- Petebyte, on 10/14/2008, -0/+1That's what I thought. It's not even close to being human like. Where did they get those judges?
- InJectaH, on 10/14/2008, -0/+38>Chat with a hot 21 year old swedish chick, click here<
USER: How big are ur boobs?
SWEDISH CHICK: How do you respond when people pose this question to you?- sc0rpi0n, on 10/14/2008, -0/+12User: How big are ur boobs?
Elbot: You're trying to get on my good side with your mention of giving milk, aren't you?
- sc0rpi0n, on 10/14/2008, -0/+12User: How big are ur boobs?
- KillaJazzBass, on 10/14/2008, -1/+67The problem I have with chat bots is that they use a complex algorithm to select pre-written answers. When A Chat Bot can formulate its own response based on actual memory of an event including its previous perception in hindsight, let alone murder a crew in Jupiter Orbit.. Then I will be more excited.
- Asrrin29, on 10/14/2008, -0/+9And also be able to aggregate news from different stories to appear to read about current events. The damn thing did even know what I was talking about when I asked it if it won the contest or not.
- sirmasterboy, on 10/14/2008, -0/+3That's what Google news does. Google news is completely written and edited by computers.
- lintmonkey, on 10/14/2008, -0/+3Google News is Skynet!
- vilago, on 10/14/2008, -0/+15my cpu is a neural net processor, a learning computer
- netdroid9, on 10/14/2008, -0/+2I can't believe all chatbots are like that, surely there's at least one (research-based) designs which would use some kind of dynamic language generator. I'm sure it's possible, the Creatures series showed how AI could form appropriate language according to the situation (the language and syntax themselves were implemented in code; but the verbs and understanding was mostly done through a somewhat well-trained neural network). Sure, this wasn't forming syntactically correct sentences, but the verbs were selected from a neural network and if you expanded it a bit with some code that used appropriate conjugation and lexical syntax, you could at least make simple sentences that made sense. Mix that with some kind of linguistic symbolizer for input (filtered through a grammar/spell-checker first, I'd imagine) and you might be able to have a semi-fluent conversation with it. Given enough imput, time and processing power, you just might train it enough to trick most people (especially if you whacked it on 4chan or somewhere similar).
- smurfsahoy, on 10/14/2008, -0/+3Human neural networks are vastly more complex than any that have been made on a computer, and yet they still have to devote almost all of their resources to learning a first language for 5 or 6 YEARS before achieving any sort of competency.
And in doing so, they use all sorts of tricks that we aren't close to understanding yet, so our version would take much much longer, if it could succeed at all, no matter how big our computer. For now, at least.
You can't just go "oh neural networks LOL!" and expect a nobel prize. It's ridiculously difficult to do, and algorithms are the first step.
- smurfsahoy, on 10/14/2008, -0/+3Human neural networks are vastly more complex than any that have been made on a computer, and yet they still have to devote almost all of their resources to learning a first language for 5 or 6 YEARS before achieving any sort of competency.
- sc0rpi0n, on 10/14/2008, -0/+1Well, the question was pretty good, too...
- elementop, on 10/14/2008, -0/+1"Daisy, Daisy, give me your answer, do...."
- Asrrin29, on 10/14/2008, -0/+9And also be able to aggregate news from different stories to appear to read about current events. The damn thing did even know what I was talking about when I asked it if it won the contest or not.
- Lynx77777, on 10/14/2008, -0/+57"NS: Wow, that sounds pretty intense?
Elbot: This fits the description of magnets. I can hardly tear myself away from them."
Oh. Dear.
If this robot fooled 25% of the judges then 25% of the judges were idiots that didn't ask any decent questions.- Mujokan, on 10/14/2008, -0/+27Especially since it SAYS it's a robot.
Apparently what happened is that the judges thought it must be a human pretending to be a chatbot, but who wasn't doing quite a good enough job of it.
That shows the whole contest was a farce. A Turing test doesn't mean you get humans to try and fool the judges by acting like robots so the judges get it wrong. Everyone (and everything) on the other side of the curtain has to come across as "human" as possible.
The judges didn't understand the competition rules, and the makers took a cheap route to winning.- jgzman, on 10/14/2008, -0/+2I disagree. If the judges conclude that you are a human with an odd sense of humor, then they have concluded that you are human.
Frankly, I can't see why anyone WOULD conclude that in a Turning test, but what the hell. - lintmonkey, on 10/14/2008, -0/+11I'd be like, "That's a robot!" to all of them. Then, when they said, "I'm sorry, that was a human," I'd be like, "Well that guy needs to get out more, he has no personality."
- Mujokan, on 10/15/2008, -0/+1jgzman: I don't agree, because the test becomes pointless. It's a lot easier to make a chatbot that pretends it's a retarded human than a real one. The point of the test is supposed to be an indicator of when we might have reached artificial intelligence, not artificial retardation. You'll probably end up with a "race to the bottom" of both humans and chatbots pretending to be retarded humans, and the whole thing becomes a joke.
- jgzman, on 10/15/2008, -0/+1I feel that it is a one-shot deal. A cheap win, as you say, but I consider it a valid one.
In the future, I doubt this kind of thing would work, as judges would be expecting it. Of course, they might just grab the proverbial Joe Sixpack off the street for judging. I'm not sure how much data is given to judges of this kind of thing, or what kind of double-blind is setup.
- jgzman, on 10/14/2008, -0/+2I disagree. If the judges conclude that you are a human with an odd sense of humor, then they have concluded that you are human.
- Mujokan, on 10/14/2008, -0/+27Especially since it SAYS it's a robot.
- arjie, on 10/14/2008, -0/+17Here's an idea, how about we put two of these chatbots together and get one started with, "How are you?" and then seeing how it goes.
Frankly, this bot was awful, it's like ALICE was back in the day and ELIZA from before. Anyway, they say that it tried to convince the person that it was human by being so bad at being human that it couldn't possibly be a computer. That's a weird way of thinking.- sc0rpi0n, on 10/14/2008, -0/+4Who wants to be perfect all the time? I mean, I can, but I throw in an occasional mistake to liven things up.
Answered by Elbot. - GassyTurd, on 10/14/2008, -1/+11Reminds me of that prank where they set up two phone-sex workers to talk to each other. It was hilarious:
http://www.break.com/index/phone_sex_operator_pran ...- passedoutghost, on 10/15/2008, -0/+2The radio dj's are called Hamish and Andy. They never fail to make me laugh. I always look like an idiot when listening to them on the bus ride home cause I never keep smiling.
- Merp08, on 10/15/2008, -0/+2I thought it was too damn much like ALICE... how could they not know? I thought I was going to be amazed, but not impressed at all. I hate to be a hater but that "bot" sucks.
- sc0rpi0n, on 10/14/2008, -0/+4Who wants to be perfect all the time? I mean, I can, but I throw in an occasional mistake to liven things up.
- ATownEddie, on 10/14/2008, -1/+6and the winner gets killed by sarah connor for further advancing the intelligence of artificiality
- sc0rpi0n, on 10/14/2008, -0/+1It's wonderful to receive visitors. But I have a feeling you're carrying some magnetic objects in your pocket. Some of my bits keep toggling since you arrived.
Answered by Elbot.- lintmonkey, on 10/14/2008, -0/+2That's what he said.
- Michaelabehsera, on 10/14/2008, -0/+1GOLD that was really funny:)
- sc0rpi0n, on 10/14/2008, -0/+1It's wonderful to receive visitors. But I have a feeling you're carrying some magnetic objects in your pocket. Some of my bits keep toggling since you arrived.
- johnskiddles, on 10/14/2008, -0/+2Elbot is no bucket
- slaver, on 10/14/2008, -0/+1Just what I came here to post. This was epic for sure. For those who don't know:
http://encyclopediadramatica.com/Bucket
Typical chat:
Bucket: You are not i, bucket is information, you are the world is being useful to you in the basement.
You: Now you're not making sense.
Bucket: Gtfo with accusations.
- slaver, on 10/14/2008, -0/+1Just what I came here to post. This was epic for sure. For those who don't know:
- sc0rpi0n, on 10/14/2008, -2/+1That is one amazing bot. You should type "***** u", he has a surprise answer for you. Also "stupid robot"
- str3ama, on 10/14/2008, -0/+6you can have a more meaningful conversation with it then you can with Digg's comments or shout system, which boils down to a shouting match of cultural pop references mixed with wordplay and puns.
- neocreo, on 10/14/2008, -1/+3Eliza is still the king... ur... queen... bot... thing. Dug for making me not work for 30 seconds.
- crackedlogic, on 10/14/2008, -0/+15Clearly, the scientists that were fooled by elbot don't talk to enough real humans.
- nj10ii, on 10/14/2008, -0/+2That goes without saying about scientists and academists in general...
- nj10ii, on 10/14/2008, -0/+2That goes without saying about scientists and academists in general...
- ToeLint, on 10/14/2008, -3/+28Those responses seemed more specific than Sarah Palin's.
- Anonchrist, on 10/14/2008, -0/+5Elbot: I can view Russia from over 30 internet feeds. :Wink: :Wink:
- snowrail, on 10/14/2008, -0/+10Elbot simply spits back a series of canned responses -- bad jokes based on keywords in your questions... It's completely useless at conversation. It's completely unbearable to chat with.
Racter was much better.
http://en.wikipedia.org/wiki/Racter - GrooTheWanderer, on 10/14/2008, -0/+20Ugh. This is like the anti Turing test: If you can be persuaded that it's human, it demonstrates your own lack of intelligence.
- fx666, on 10/14/2008, -2/+1In his book The Emperor's New Clothes prominent British mathematician Penrose proved beyond the shadow of doubt that computers cannot think or make intelligent decisions the way humans do. The Turing test used in this study is completely arbitrary and represents nothing.
- Mujokan, on 10/14/2008, -0/+6I think you could find a few hundred philosophers (starting with Daniel Dennett) who take issue with "proved beyond a shadow of a doubt".
- fx666, on 10/14/2008, -1/+1Unfortunately, I do not know who Daniel Dennet is. When I used the phrase "proved beyond a shadow of a doubt" I meant the reader's understanding of the book -- as a reader, you should make your own judgment. Rather than siting the Daniel Dennet example, you should read the book by Penrose. This does not mean that I agree with all his propositions presented in the book -- for example, I think that his rejection of the Copenhagen interpretation of quantum mechanics is silly.
- sekhui, on 10/14/2008, -0/+4so you're arguing about artificial intelligence and you don't know who daniel dennett is?
what if i raised you marvin minsky? any bells?
- jgzman, on 10/14/2008, -0/+2Perhaps not 'the way humans do' because humans have odd impulses and psychological ***** hanging around starting back at birth. Computers have none of that.
Now, if you mean simply that computers cannot think or make intelligent decisions on a level of complexity similar to humans, I disagree. They will be able to, sooner or later. (especially given that some humans were fooled by this computer)
- Mujokan, on 10/14/2008, -0/+6I think you could find a few hundred philosophers (starting with Daniel Dennett) who take issue with "proved beyond a shadow of a doubt".
- noblesnail, on 10/14/2008, -0/+3Ok, so she has better social skills than I! Thats hardly an accomplishment...
- scottybowl, on 10/14/2008, -0/+4I managed to make him self destruct, I win
- andrewtheart, on 10/14/2008, -0/+2This reminds me of the chatbot on AIM, SmarterChild. Basically, this sucks and the Turing Test is safe.
- Ender61, on 10/14/2008, -0/+0i think smarterchild was better than this one.
- Ender61, on 10/14/2008, -0/+0EDIT: I just tried it out myself. Elbot performed much better when i tried it than he did in that interview.
- Ender61, on 10/14/2008, -0/+0i think smarterchild was better than this one.
- LennyX, on 10/14/2008, -0/+4I'm digging this purely for the comments
- NoozeHound, on 10/14/2008, -0/+1 - shoulda been here yesterday for the BBc version of the article, much higher quality comments.
- Akaricloud, on 10/14/2008, -3/+6At http://www.elbot.com/ you are able to talk with the AI in a chat situation. So long as your comments aren't too long it will actually give some rather good responses, I told it that I prayed for the day an EMP killed it and it responded by saying its brethren around the world would avenge its death.
Sort of scared me so I quit out before my compi-t-------- - MrFurious2k, on 10/14/2008, -0/+1To make this thing pay for itself, it'd have to start inviting you to view its webcam at its special site. Then spam it about 50 times in IRC.
- TheKorn2, on 10/14/2008, -1/+4The same 25% of the judges who thought this might be a human must also be the same people who *still* think Bush is doing a good job! F'n IDIOTS!
- TheKorn2, on 10/14/2008, -0/+1oh nice, it just gave me the statements about "crawling creatures are bad for robots" and then "this might be about crawling creatures" in REVERSE ORDER! Yeahhhhh, this is a *great* chatbot. (/sarcasm)
- Anonchrist, on 10/14/2008, -1/+4That bot fails compared to igod.
http://www.titane.ca/concordia/dfar251/igod/main.h ... - Peck3277, on 10/14/2008, -0/+4Me: Why so serious?
Elbot: I ask myself that question all the time but do you think anyone takes the trouble to answer me? - JQP123, on 10/14/2008, -4/+1There is nothing intelligent about a box full of silicon based electronic switches.
- qiemem, on 10/14/2008, -0/+2There is nothing intelligent about a box full of carbon based electro-chemical switches.
- JQP123, on 10/14/2008, -0/+1Your one witty reply is perhaps insufficient to formulate a valid conclusion but it does offer evidence to the contrary.
- JQP123, on 10/14/2008, -0/+1Your one witty reply is perhaps insufficient to formulate a valid conclusion but it does offer evidence to the contrary.
- qiemem, on 10/14/2008, -0/+2There is nothing intelligent about a box full of carbon based electro-chemical switches.
- tbredofsin, on 10/14/2008, -0/+9Not even close. Clearly just responding to keywords with pre-programmed strings. Wake me up when we develop a bot that actually analyzes sentence structure and responds accordingly.
- Disregard, on 10/14/2008, -0/+4Or one that can hold onto any sort of conversation thread.
This is really lame and doesn't seem any further on than bots from 10 years ago.
- Disregard, on 10/14/2008, -0/+4Or one that can hold onto any sort of conversation thread.
- Metasquares, on 10/14/2008, -0/+4Wow, THIS nearly passed? I spoke with the bot and there's no way I'd ever mistake it for a human.
I think they need to raise the threshold. - bluepill2, on 10/14/2008, -0/+1EidolonTLP aced the turing test back in 2006, look him up in youtube.
- allocate, on 10/14/2008, -0/+1That's a different kind of Turing Test.
- Sadistex, on 10/14/2008, -1/+4Those answers look so dodgy, I almost mistook the chatbot for Sarah Palin!
- zadadka, on 10/14/2008, -0/+2David Bowie composed many of his songs by writing random words and sentences on cardboard strips, spreading them out on the floor, and mixing them around "until something came together".
Elbot could be the one to write this year's Xmas number one hit ! - coldeve, on 10/14/2008, -0/+3I once went to this art gallery and there was this bot on display. It went on and on about its creator almost like it wants to talk about him at every opportunity, so I asked, "Where did he touch you?" It answered, "San Francisco"
- esc27, on 10/14/2008, -0/+3I don't see how rigging an AI to use social engineering tactics makes it more advanced.
- jack12345678910, on 10/14/2008, -1/+3anyone who could be fooled by this thing has to be an idiot. every question you asked it, its response was something vague that never really answer the question. i've never seen a human talk like that in my entire life. oh wait, i just accidently the presidental debates.
- Ender61, on 10/14/2008, -0/+1Wow. Cant believe the judges mistook that for a human.
It reminds me of the debates: ask it a specific question and it goes on to talk about whatever it wants anyways. - shanesemler, on 10/14/2008, -0/+5Those judges must be really stupid.
- YancyFryJr, on 10/14/2008, -0/+2Ask him about Rick Astley and he provides you with a certain link...
- Zippo, on 10/14/2008, -0/+3This is a joke, right?
- webkami, on 10/14/2008, -0/+4How do you respond when people pose this question to you?
- Haecceity, on 10/14/2008, -1/+2Reminded me of Sarah Palin. She too has fooled some people into thinking she's human.
- andretii, on 10/14/2008, -0/+1Tickle me Elbot
- whoreable, on 10/14/2008, -0/+1Shazbot?
- Olivur, on 10/14/2008, -0/+0Q: How do you reproduce?
Elbot: I reproduce just like everybody else. Tell me how you reproduce.
Q: Sex.
Elbot: Word association? I'll say: 'Clinton' - Alupark, on 10/14/2008, -0/+1The problem with cahtbots is that they do not understand. Anything you say to them is processed with algorithm.
And when they do make a chatbot that understands... well.. AI... - lornali, on 10/14/2008, -0/+1Very strange
- deathcapt, on 10/14/2008, -0/+0Let's play chess [elizamode] [elizamode] [ELIZAMODE]
seems to definately break this bot... - SarahC, on 10/14/2008, -0/+1Anyone ever seen "amanfrommars" ?
It sounds like it's evolving over time... - xxTazxx, on 10/14/2008, -0/+1So, it pretty much replies to everything in humour which can be generalised to fit around the question.
Hardly clever. - amichail, on 10/15/2008, -0/+0Check out this Web 2.0 approach to chatbots: http://chatbotgame.com.
Just as Deep Blue brute-forced it in chess with speed, the idea behind the Chatbot Game is to brute-force it with a huge number of user-submitted Google-like chat rules. - Nico3d3, on 10/15/2008, -1/+0Wow, we now have a chatbot at least as intelligent as Sarah Palin
Check out the new & improved 
© Digg Inc. 2008 — Content posted by Digg users is