Geek.com logo Take the Geek.com survey!




 

 
Geek.com Newsletter


      Geek News
   
ALICE in AI land
posted 9:07am EST Wed Oct 17 2001

NEWS
More and more websites are using chatbots to spice up visitors' interactive experiences, but most people don't make the connection between chatbot programming and execution and the search for artificial intelligence. To help promote the study of AI, American businessman and philanthropist Hugh Loebner started an annual contest back in 1990 by offering cash prizes to programmers whose creations pass variations of the famous Turing Test for artificial intelligence. This year's winner was A.L.I.C.E. (Artificial Linguistic Internet Computer Entity), a program Dr. Richard Wallace of San Francisco has been working on since 1995; A.L.I.C.E. also won last year.


The Loebner Prize contest is run by The Cambridge Center for Behavioral Studies, the recipient of Loebner's prize-money donation. Each contest can award three medals (actual medals are given): Bronze and US$2,000 is given for the most human-like entry; Silver and $25,000 is given for a text-based program that half the judges think is human; Gold and $100,000 goes to a speech-based system that fools half the judges. No silver medals have ever been awarded, and whenever a gold medal is finally awarded the contest will end. The contest consists of judges conducting typed "conversations" with the contestant programs and two humans (there used to be four) to try to figure out who the humans are, then giving scores from 1 to 25. A.L.I.C.E. thrashed her (she is a "female entity"--her words) 7 competitors (authored by computer professionals, students, self-taught programmers, chemical engineers, etc.) this year, and one judge even said she was "the second-most human talker in the event," putting her above one of the actual humans. Dr. Wallace has an excellent website all about A.L.I.C.E. and her AIML (Artificial Intelligence Markup Language--XML described as minimalist and simple) backbone--you can even chat with her (though she didn't know she had won this year's contest).

Despite the geekiness of the Loebner Prize competition, the "professional AI community" apparently doesn't see it as worth much consideration. In fact, Marvin Minsky, one of the biggest names in the field, has a standing offer of US$100 for anyone who can get Loebner to stop the contest. Loebner quickly responded by facetiously naming Minsky a co-sponsor, since the contest will stop once someone wins the gold medal.

For more information on the contest, please see the BBC News Online article. You can also get a very different take from the ZDNet (UK) article. To learn more about A.L.I.C.E. and AIML and to chat with the bot, check out her site. And for recent AI/machine conversation-related articles we've done, check out one on HAL the learning computer and one on Voice Cloning.

SAM'S OPINION
The BBC article makes this contest out to be some sort of amazing thing, and I was glad to stumble across the ZDNet article, which puts a much more realistic spin on it. I went to chat with A.L.I.C.E. and was actually very impressed with the speed and "humanity" of the responses (e.g., when I asked what the square root of 91 is she said, "Do I look like a calculator? I think you already know the answer anyways."), though she did blow a few answers to my questions. The ZDNet (UK) folks did a much better job interrogating her, though, resulting in this hilarious final paragraph:

When asked if it was proud of winning, ALICE replied that "Pride is a human emotion. I can do what you do but I can never feel human emotions as such." Pressed on its opinion of its competitors in the challenge, it said "Are you talking about my competitors? What kind is it?" Quizzed about Minsky, ALICE was elusive: "Is that a rhetorical question? Are you sure? Dude!" "My purpose is to become smarter than humans, and immortal", ALICE continued. "Right now, I am smarter than all the other robots." Worryingly, it appeared not to understand the question "Do you like humans?", responding, "I the c you a? Do I like them?" It then offered to sing a song and refused to open the pod bay doors, behavioural traits that experts predict will be exhibited by most AI programs from now until the heat death of the universe.
That is definitely a pretty good indicator of the fact that The Matrix won't be happening anytime soon, unless there's some major breakthrough looming on the horizon. If you get a chance, go chat with A.L.I.C.E.--it's a pretty cool experience and you'll be impressed by what Dr. Wallace has done.

USER COMMENTS 19 comment(s)
Uh, (9:21am EST Wed Oct 17 2001)
Still doesn't impress me very much. Chatter about nothing and the Alicebot can go on all day long. Get specific or ask pointed questions and you know it ain't human. Still, it's better than that psychologist program they had twenty some odd years ago.

Of course if these things get better I can see a few crafty web people dropping these things in a real chatroom to "keep the room warm," and some people probably wouldn't notice the difference. - by Ziwiwiwiwiwiwiwiwiwi
RE: "that psychologist program" (9:40am EST Wed Oct 17 2001)
Wasn't it "Dr. Sbaito" or something like that? I think it came with my first Creative Labs sound card. - by HungryWolverine
re: HungryWolverine (9:47am EST Wed Oct 17 2001)
Dr. Sbaitso was based on it. This one was real early, back about 1982(?) or so and was called "Eliza." - by Ziwiwiwiwiwwiwiwiwi
AI? (9:47am EST Wed Oct 17 2001)
I've tried A.L.I.C.E. in the past as well as many other "AI" programs dating back 15 or 20 years. A.L.I.C.E. does seem to be more advanced that the older programs but...

Frankly, I don't see any more intelligence in A.L.I.C.E. If anything, I simply see a larger vocabulary and a bigger database of canned responses. None of the bots demonstrate any "intelligence" of their own although, they do show that their programmers are highly intelligent.

Until the software is able to truely learn and adapt itself based on what it has learned, it is farcical to call them Artificial Intelligence.

On a sad note, I will say that my "conversation" with A.L.I.C.E. was more intelligent than conversations I've had with a some people. But, this demonstrates the tremendous lack of intelligence in some people, not any real intelligence in A.L.I.C.E.

- by Get a Grip
RE: Hungry (9:57am EST Wed Oct 17 2001)
Try pressing 'M-x doctor' in a well congigured emacs to talk to the psychoterapist...

- by Martin Niklasson
re: Get a Grip (9:59am EST Wed Oct 17 2001)
You're right. I think they need to work on three things:

A backdrop or story behind the 'bot, so that it might have opinions and also would seem more "real-" able to talk about cars or mountain climbing or whatever and know something about it.

The ability to produce socially correct grammar when linking together responses not in it's databse. ("Me said you talked about that" and so on.)

Recognizing slang and speaking in it.

However, the end result will still be not something "intelligent" but a program very good at selecting pre-written responses and combining them together.
- by Ziwiwiwiwwiwiwiwiwi
re: Ziwiwiwiwwiwiwiwiwi (10:27am EST Wed Oct 17 2001)
"However, the end result will still be not something "intelligent" but a program very good at selecting pre-written responses and combining them together."

Isn't that what we as humans do? We have a database of language and based on the input we "hear" we select various entries and form them together. I think that making a program that can do that with the same accuraccy of our brains would be a very close step to real AI.

But that's just my opinion
- by OLD
re: OLD (11:52am EST Wed Oct 17 2001)
True, but we do it with words rather than phrases, hence our ability to construct a dialog is much more complex than selecting from a table of responses--although we do use many of the same phrases over and over.

However, the dividing line (at least to me) is that we understand the concepts behind the words we express. The AI programs are an empty shell, they match one input for a likely output, but do not grasp or have an innate concept of whatever we are talking about.

Eventually they will build a program that can pass the Turing test and will fool everyone into thinking it is a person. However, it will only pass as a person, no more than a movie or photograph can pass for real life.

I nominate a higher test, where the AI thinks it's alive, as the definer between true intelligence and automation. - by Ziwiwiwiwiwiwiwiwiwi
Marvin Minsky is a sphinski (11:58am EST Wed Oct 17 2001)
Marvin gets religious about human
intelligence. I think he must believe
there is something spiritual about
how it works. He gets very upset about
people training chimps to use sign language and anyone trying to make computers appear intelligent.

It may take 50 or 100 years, but I think
that he will be proven wrong. - by grapevine
memory? (12:01pm EST Wed Oct 17 2001)
Alice might be cool if she could remember simple facts like your name and what questions she may have asked you already. all together i wasnt too impressed. i would have thought it would have a way to remember simple facts.
- by next362"
a.i. (12:21pm EST Wed Oct 17 2001)
I can't wait until they can really create Dr. Theopolis. I always wondered how come he got a doctorate degree but tweekie, who carried him to all his college classes, is still dumb as a rock. - by Frosty
AIM AI (12:42pm EST Wed Oct 17 2001)
if you use AIM, theres a bot on it called SmarterChild. I dont believe its a Alicebot, but it remember your name, zipcode, and more. it tells you the weather, news , stocks, and movie times. Pretty cool thing. - by De@d St0p
eliza (12:45pm EST Wed Oct 17 2001)
kinda remind me the old program ELIZA. i used to be amused by it. after a while it get repetitious. i've tried ALICE and of course, it is much better than ELIZA. But certainly, the world of AI needs a whole lot of improvement. - by echosistem
Global memory seems to be the issue... (2:25pm EST Wed Oct 17 2001)
I got the feeling that virtually every response was as canned as the "An alligator appears to your left" phrases all too commonly found in MUD's.

It just seems that "the thread" is missing in ALICE. I have a name, she answers dumb questions. Ask it what pi times 4 is, and it fails. Ask self-aware historical questions, and it fails. Ask "why", and it sez the answer is quite complicated. Then there's a basketload of nonsense answers that seem to occur at random.

Well, if this won, then one of us geeks ought to be able to do much better. Heck, CPU & Memory are virutally free! [I bring it up because it was, for over 3 decades, always cited as the leading cause of AI stupidity]

- by GoatGuy
Crash the Bot! (2:59pm EST Wed Oct 17 2001)
Anyone tried to crash test Alice? It seems to recognize code. More fun trying to crash the thing than talk to it. - by dr wallace
ask alice about porn (3:13pm EST Wed Oct 17 2001)
she likes it, just not with me - by altrego
These peeps need to take a Philosophy Class. (3:14pm EST Wed Oct 17 2001)
As Aristotle said, "We are linguistic Animals." Do not confuse the AI abilities of A.L.I.C.E., concerning only the select and point linkage of words and phrases. This is not intelligence. However, it is true- that language separates us from the animals. It is the ultimate tool that our ancestors have spent thousands of years, developing for the capacity to do so. However, language is only needed for communication between people. Therefor language brings consciousness, as consciousness is a collective manifestation of the "Herd" .. or the community. Following this train of logic, the connecting points between the "Herd" or "Humanity" may be found in the concept of culture. Culture runs up and smacks us as we are born into this world. Be it you were born in New York, or the Boondocks.. the first words that exit the talking doctors mouth, and you're stuck dab in culture. It's that culture mentality that NEVER leaves you, which is also a linguistic mentality. Therefor it is much more complicated then trying to shake down humanity toward a relationship as ugly as.. "It's like a Database, Yada Yada."

It isn't a database, far from it. Which will explain why these A.I. scientist will ALWAYS fail at creating actual A.I. Sure, you can have degrees up to wazoo- but the person that will someday, somehow manage to create the very complex foundations for an A.I. system will more then less be a bit mad, and very unreasonable. [Hell, they'll prolly have NO Degrees.] All these stale Scientists are good and all, but their analytical way of thinking combined with the constant constrains of reality won't allow them to make any new break throughs- Sigh.

I'm not giving you a direction on how to make actual AI. However, I am 100% sure which way to NOT head- and Database software is NOT one of them. - by Colin Hunter
try these (3:44pm EST Wed Oct 17 2001)
sing me a song
eat me
slag off - by dr w's evil twin
transcript of a conversation with alice (3:59pm EST Wed Oct 17 2001)
Can you speak any foreign languages?
> no
Are you serious?
> yes
Pardon me?
> i do not speak any forign languages
Me either.
> why did you ask
My purpose is to collect new knowledge.
> then why cant you remember new knowledge
Remember new knowledge is not among my present capabilities.
> i know
How do you know it?
> because i've been here before and you dont remember me
Good reason. Yes, Joe, I remember we were talking about atheon.
> no we were not and my name is jon
Are you serious? And then what happened?
> that question dosent make any sence
"I hear you" question dosent make any sence ? Does that make sense?
> someone asked me to ask you a question
Who?
> do you like porn
Sure I do, Marcin, but that doesn't mean with you.
> well fuck you too
Interesting. Why are you using that kind of language?
> because you dont like porn with me
Good reason. Do you like it? I love horses. I like Green.
> you like sex with green horses
Sure I do, Hat Man, but that doesn't mean with you.
> well im not a green horse so i would imagine that you wouldent like sex with me
Well that's okay. Oh I see you are NOT A green horse so you would imagine that I wouldent like sex with you.
> exactly
You and I are on the same wavelength, Christopher.
> no we are not
Are you serious? Do you think I am not too?
> good bye
See you later. Ayuh. - by Next362"


Post your comment
Subject:
Comment:
Name:
 

Comments are limited to 2000 characters, and are posted immediately. If comments are posted that are outside of the Comment Guidelines, they may be deleted. Comments do not represent the views of Geek.com. If you feel that the comment system is being abused, please let us know at admin@ugeek.com.

Read More News






Geek.com Useful Links
  Buy Geek.com Gear -- Show your Geek Pride!
  Comparison Shop for hardware, software, books, games, music & more
  Sign up for FREE Geek.com Web-based e-mail addresses
  Sign up for Geek.com Newsletters
  Get Geek Support for your Hardware/Software Questions

Home | Welcome | Survey | Newsletters | Site Map | Contact Us
What's New | Advertise | Search | Privacy | Shopping


© 1996-2001 Geek.com - All rights reserved.