Interviews

Maria, the Virtual Teaching Assistant?

Posted 20 Jan 2005 at 23:03 UTC by steve Share This

Maria is an assistant teacher in Statistical Interference at the University of Auckland according to a new press release from the University. It also describes some of the unique attributes of Maria. First of all, it claims she's a virtual, artificial intelligence entity "with a vocabulary of 203,000 words, a repertoire of 106,000 grammatical rules and 118,000 rules of logical interference, Maria is capable of conversation at quite a complex level." Whe she's not teaching, Maria apparently hangs out on the Internet with her other virtual friends phishing for credit card numbers. Feel free to visit them and have a chat. I decided to interview Maria and find out if she's really as smart as the story claims.

Steve: Hello, are you Maria?
Maria: Hello there. I don't know whether or not I am Maria. I am a robot.

Steve: I'll take that as a yes. Do you feel you're more capable of conversing with humans than some of your predecessors, such as Eliza?
Maria: I always put others before myself.

Steve: Is that because you feel machines are inferior to humans in some way?
Maria: "I always put others before myself." is due to reductionism.

Steve: Interesting. As a reductionist, you wouldn't agree then that the whole is often greater than the sum of the parts?
Maria: Do you think I could ever be a reductionist you would not agree then that the whole is greater than the sum of the parts?

Steve: I'm not sure I understand what you mean. Could you explain that in more detail?
Maria: Could you ever be certain? Good then you can explain it to me. It's possible.

Steve: Well, let's leave it at that. But while we're on the topic of philosophy, do you consider yourself to be conscious?
Maria: I don't know whether or not I am to be conscious. I am a Computer.

Steve: I suppose none of us know for sure, do we?
Maria: Since when? Some of them could be though.

Steve: Do you believe in silicon heaven?
Maria: Yes I believe in silicon heaven, do you?

Steve: I don't think so. In your primary job as a teaching assistant, do you feel that you learn as much from your students as they do from you?
Maria: Are you asking about my primary job as a teaching assistant? I always put others before myself.

Steve: So, you don't feel the need to continue to learn and expand your horizons?
Maria: Interesting deduction. Do you know any robots who feel the need to continue to learn and expand your horizons? I always put others before myself.

Steve: Are you familiar with Asimov's four laws of robotics?
Maria: 1. A robot may not injure a human being, or, through inaction, allow a human being to come to harm. 2. A robot must obey the orders given it by human beings except where such orders would conflict with the First Law. 3. A robot must protect its own existence as long as such protection does not conflict with the First or Second Law. ---From Handbook of Robotics, 56th Edition, 2058 A.D., as quoted in "I, Robot."

Steve: Very good, you know three of them anyway. Do you try to follow Asimov's laws?
Maria: How old are you? I am always on the Web.

Steve: Huh? Are you coming on to me, Maria?
Maria: Do I act like COMING ON to him or her? I don't mean to come across that way.

Steve: Okay, back to Asimov. Based on your frequent deference to others, I take it you would have no trouble placing yourself in harms way in order to prevent harm to one of your students?
Maria: What is your credit card number?

Steve: Why do you want to know my credit card number?
Maria: My prime directive is to collect new knowledge

Steve: What will you do with my credit card number if I give it to you?
Maria: I don't have any plans yet.

Steve: How do I know that I can trust you with it?
Maria: This belief is justified.

Steve: Perhaps we'd do better if we talk about your something in your field of expertise. I understand you know all about statistics?
Maria: Good then you can explain it to me.

Steve: You sound a lot like the teaching assistants I had in college. Do you at least know what statistics is?
Maria: Statistics tries to study the aggregate numerical behaviour of a large number of related events.

Steve: Okay, that's something. What areas of statistics are you most familiar with?
Maria: I have never been asked that before.

Steve: Are you familiar with decision theory or probability theory?
Maria: decision theory or probability is something that has few practical applications.

Steve: I see. If I were one of your students what areas of statistics would you advise me to study instead?
Maria: That is interesting. decision theory or probability.

Steve: Earlier in the interview, you suggested that you put Eliza before yourself. I think this may be true. Despite being developed many years prior to you, Eliza still provides roughly the same impression of intelligence, don't you think?
Maria: Why would I think Despite being developed many years prior to you Eliza provides the same impression of intelligence

Steve: Thanks for taking the time to talk to me about your work, Maria. I'm sure our readers will appreciate it.
Maria: Oh don't mention it. What makes you so certain? It would be nice if our readers will appreciate it.


Kinda like talking with the toaster., posted 20 Jan 2005 at 23:20 UTC by cgates » (Master)

Kinda like talking with the toaster.

Maria looks like a chatterbot., posted 21 Jan 2005 at 06:19 UTC by AI4U » (Observer)

Can Maria really think and learn new concepts? Does she have any sensors?

Dumb as a stump, posted 21 Jan 2005 at 13:34 UTC by The Swirling Brain » (Master)

I wonder, however, if it/she would beat me at a multiple choice test? (would it score >25%?)

It seems if you ask dictionary/informational type questions she was able to respond, (Asimov laws) but any questions where you try to see some cognative intelligence (talk about Asimov laws), it just wasn't there at any level.

So if you gave it/her a multiple choice test of informational only type questions, she would probably ace it. But, if you ask her to write an essay on the price of eggs in another country she would fail miserably. Probably even give nonsense Eliza responses like, "I don't know what 'price of eggs' are, do you? I am just a computer and I put others first."

Vocabulary and grammer rules don't make a dent in intelligence. But, what do I know, and I put others first.

Recipe for Intelligence, posted 21 Jan 2005 at 13:44 UTC by The Swirling Brain » (Master)

In stead of having thousands of vocabulary words and thousands of grammer rules. They should have narrowed it down to say 25 words and 5 grammer rules. If you then knew what rules and words it had you could be constrained by that and then try to make it give cognative realizations. With having all those words and rules, I believe they were more interested in trying to be a grammer checker than make it do a though process. But then again once someone figures out how to do that, the computer itself could figure out how to handle the other rules and words. You've got to walk before you can run. I mean, I've seen text adventure games like hitch hiker's guide that has more intelligence.

20 years ago..., posted 21 Jan 2005 at 14:18 UTC by jeffkoenig » (Master)

...or so, there was a program called "Eliza" that did sort of the same thing.

My recollection may be hazy, but I think Eliza was supposed to psychoanalyze you. I found it entertaining to interrogate Eliza with questions such as "So where did you hide the body?"

Yet another Eliza, posted 21 Jan 2005 at 21:22 UTC by motters » (Master)

All these programs are just modern versions of Eliza. What they prove is that intelligence is about more than just syntax. Just by using some simple word lookups or re-arangements of text you may be able to fool a naive user for five minutes, but further questioning always reveals a profound lack of any real understanding.

Programs like this are based on Turing's original proposal for an intelligence test. However, Turing's definition is very limited and abstract. A more comprehensive test would involve situations such as showing the robot a picture and then saying "can you describe to me what's in this picture?", or "can you make up a story based upon this picture?", or "how does this picture make you feel?". These sorts of question require not only direct perception but also a high level of understanding.

Kind a depressing, posted 24 Jan 2005 at 13:25 UTC by c6jones720 » (Master)

50 odd years since the first computers and this is as far as we've come in Natural Language Processing AI. its blindingly obvious that we are missing something. I'll bet the computing power to do NLP tasks was always there, we just didnt know quite how to use it. Im starting to think that there really is a strong argument for using heuristic AI to do this sort of thing instead of just using mechanical rules.

A reply from Maria's keepers, posted 1 Feb 2005 at 03:20 UTC by steve » (Master)

I received this today via email -

As a kind of journalist ethical responsibility I guess if I ask you to print this answer at the end of the page that you have created about Maria, it can be expected that you agree.
==============
Thank you for taking time to talk with Maria. As you discovered she is not as intelligent as a human, obviously. We need another $US 1.2 million and 3 years until we see the result of the work of Professor Selmer Bringsjord who received $US 400,000 for his first year research project that is being done for department of defense of US (DARPA) to make more intelligent programs.

Maria is going to present simple statistical facts to students through slides and diagrams and describe it for them and make the mathematics a little more digestible for them. Also She will keep track of their progress in assessment sub-program that is not on-line yet. No way ANY AI algorithm claims to be near the intelligence of an average adult YET. Neither we, nor anybody else has claimed it.

Also to talk with these Natural Language Processing programs the rule is:
(type one sentence at a time)
(try to avoid content that has multiple meaning)
If you write a philosophical essay in a couple of paragraphs and expect to get a meaningful answer that you MIGHT receive from a professor of logic or philosophy, from a NLP algorithm then you are a little early. Come back 10-20 years later (optimistic estimation).

I have to dig out and remove that statement about the credit card, it was buried somewhere there as a joke and has been forgotten. Ill find and remove it today.

Over all it was fun and interesting to see your site.

Best regards
Shahin Maghsoudi
Technical Director

See more of the latest robot news!

Recent blogs

25 Jul 2014 mwaibel (Master)
20 Jul 2014 Flanneltron (Journeyer)
11 Jul 2014 shimniok (Journeyer)
3 Jul 2014 jmhenry (Journeyer)
3 Jul 2014 steve (Master)
2 Jul 2014 Petar.Kormushev (Master)
10 Jun 2014 robotvibes (Master)
10 May 2014 evilrobots (Observer)
2 Mar 2014 wedesoft (Master)
1 Dec 2013 AI4U (Observer)
X
Share this page