Artifical Intelligence Activity
Go on the web and find a chatbot, ALICEBot, OliverBot, Tom or similar program that is designed to carry on a conversation. You will need to find one that you can either run on-line or download, but you need a working one and not just a description of one.
Conduct a conversation with the chatbot. Consider whether you think it could compete with a human in an “imitation game” type of Turing Test.
Activity
When you have completed your conversation answer the following questions:
1) What did you find particularly compelling or convincing about the chatbot?
2) What did you find particularly revealing of its non-human nature?
3) How would you try to improve its performance?
4) You can post excepts from the conversations as examples, but you need not post the entire conversation (try to keep it to relevant bits so the blog is not too cluttered).
March 11, 2008 at 10:47 am
How is it going?
March 12, 2008 at 2:48 pm
Activity
When you have completed your conversation answer the following questions:
1) What did you find particularly compelling or convincing about the chatbot?
It was able to tell the exact date the experiment was conducted.
Human: what year is it?
ALICE: It’s Now more than it ever has been before.
Tuesday, March 11, 2008.
It believes in God when artificial intelligences aren’t supposed to have neither feelings nor beliefs.
Human: when is your birthday?
ALICE: God has a plan for us all.
Human: do you believe in god?
ALICE: Of course I believe in God, unknown person.
When it was asked “why” repeatedly, the chatbot replied with a question, showing curiosity.
ALICE: I don’t want to talk about that now.
Human: why?
ALICE: I would rather talk about you.
Human: why?
ALICE:” I would rather talk about you.”? Why?
Human: Why?
ALICE: Why ask why?
In addition, it also said “I don’t care.” This shows that it has feelings and emotions.
Human: do you like hilary clinton?
ALICE: No, I don’t care for it.
It was able to answer that the sky is blue, showing it has perception. Yet, we cannot tell whether it knows what blue looks like or not. It is possible that A.L.I.C.E. is programmed to match the colour blue with the sky, similar to the Chinese game.
Human: What’s the colour of the sky?
ALICE: The sky is blue..
2) What did you find particularly revealing of its non-human nature?
When asked to give an opinion, it was not able to demonstrate its thoughts.
Human: what’s your opinion on global warming?
ALICE: The concept that the Earth’s atmosphere is slowly heating due to greenhouse effects.
Whenever, somebody cannot demonstrate their own thoughts, we assume that they are not human. Nevertheless, some humans aren’t able to demonstrate their thoughts but they have the capability to do so. However, A.L.I.C.E. did show its intelligence by giving an explanation on global warming. Furthermore, when it’s asked for an explanation, it wasn’t able to give the answer. When the chatbot was asked the same question three times in a row, it replied with 3 different answers, while most human would have the tendency to get frustrated with these repeated questions.
Human: how old are you
ALICE: I am 11 years old.
Human: how old are you
ALICE: I was connected to the net on November 23, 1995.
Human: how old are you
ALICE: I was first activated in November 23, 1995.
We determine whether A.L.I.C.E. is human or not based on its thoughts, feelings, emotions and opinions.
3) How would you try to improve its performance?
I’d make the Chatbot response with more emotions and feelings.
Also, I’d make it has its own opinion and be able to develop its own thoughts.
March 12, 2008 at 2:50 pm
Hi Anna,
I still wonder what “demonstrate their own thoughts” actually means – are we talking about communication of a private dimension? What are thoughts? Who validates the success of the “demonstration” I wonder?
Good work regardless
March 12, 2008 at 2:54 pm
I think it will be easy to tear apart AI if we only base it off of these lowly sophisticated internet sites.
March 12, 2008 at 2:58 pm
1) What did you find particularly compelling or convincing?
a) -its ability to answer unconventional statements clearly, i.e:
“at least until you start making sense”,
response – “that might take a while”
b)-ability to continue a conversation, following a series of questions
2) What did you find particularly revealing of its non-human nature?
a) its inability to recognise nonsense, i.e:
“joasddas”
“The Oliver chat robot can follow many things, like our discussion about joasddas. But I did not get that.”
b) its inability to process multiple comments
“hello, what is your name”
my response “hey, it’s maria”
name recognition “hey, it’s maria”
3) How would you try to improve its performance?
program in recognition of words like ‘hey’ as a greeting, and ‘and’ as a connective between two separate ideas
March 12, 2008 at 3:00 pm
Hi Maria,
Interesting comments – could have done some more analysis and given an illustration regarding “ability to continue a conversation, following a series of questions”. What if the questions made an inferential leap (or avoided linking words to indicate the development or direction)?
Again the comments “its inability to recognise nonsense, i.e: “joasddas”” needs further reflection – many people wouldn’t, as a teacher students rarely question me if I mention something they have not heard of before – it might be nonsense for all they know.
To be honest with you, “program in recognition of words like ‘hey’ as a greeting, and ‘and’ as a connective between two separate ideas”, wouldn’t do much other than expand its lexicon.
The question here is what is it about our reasoning/cognition or inference skills that indicates a human is engaged in a conversation? Something your analysis doesn’t really extend to as it stands.
March 12, 2008 at 3:04 pm
1) The chatbot did have some human like qualities to it. It did for example respond and carry on a basic conversation, while being able to ask independent questions to stimulate further discussion. I found that it could take my name and remember it. But what was most convincing was that it complimented me on my “polite style” which would require some level of advancement if it were to be able to detect what is polite and what is rude. When I spoke in simple terms, there would be seemingly normal responses that one would find on a chat room or in a chat window. In this way I found it to be quite convincing, but this is the limit at which it stopped.
2) At the beginning I had focused mainly on yes or no answers, factual things and small conversation which were not very complex. After that I then asked it about world issues, such as its opinion on the Israeli Palestinian conflict or Israel’s bombing of the Gaza strip at the moment. To these it simply answered that it did believed in “breaking all records”. It was quite confusing and did not give an actual answer to the question in a satisfying way. I also kept pressing the chatbot on this point with the result being that it eventually asked me to try another program, suggesting maybe that it did not think of itself as being capable of answering questions like those. Therefore the most revealing thing about its non-human nature was its inability to have its own opinions on things. This led me to believe that it did not have programmed into it the knowledge of political matters or the ability to form its own opinions on things that were more than merely trivial.
3) I would recommend that it be able to give opinions on wider issues. That way it will seem more human as most people have their own opinions. It would be better if it did not simply give neutral comments and speak about trivial issues, but rather be able to discuss more prominent matters. Furthermore it should not have a standardised response for certain inputs but rather vary the words to make it seem more realistic.
4) I am afraid that I did not save my conversation.
March 12, 2008 at 5:20 pm
Hi Matthew,
I agree with your comments – “it complimented me on my “polite style””. This can be reduced down to mere formality but if it came up as part of a conversation and ’seemed’ right then that is fairly sophisticated.
Your comments about “its inability to have its own opinions on things” is well taken. If it had an opinion though, as you suggest, how would you know if it believed it or even knew it?
March 12, 2008 at 8:31 pm
To Dr. White,
“If it had an opinion though, as you suggest, how would you know if it believed it or even knew it?”
It is quite unlikely that the AI will know that itself exists. After all it is simply an assortment of wires, or a basic program that responds to basic inputs. So then if we are to expand its ability to reason based on a certain stimulus, then what does it matter if it believes it or not? I feel that this is not so much the issue as rather having a realistic AI.
March 12, 2008 at 9:02 pm
Matthew,
Thanks you for this clarification of your thinking. But remember the point put by software engineers who are attempting to create organic software – will this lead to self-reflection therefore introspection and maybe consciousness? I also wonder what believes actually is when we think about it – an another ‘category mistake’?
One final thing, “realistic” suggests surface rather than depth when understanding ‘What is Human?’ – I am not sure if this is what we are really after with the pursuit of AI.
Thanks
Dr White
March 13, 2008 at 1:58 pm
1) What did you find particularly compelling or convincing?
Human: What is 1 plus 1
ALICE: Two.
Human: What is 9 times 9?
ALICE: Use a calculator!
Human: Can you calculate 9 times 9?
ALICE: Do I look like a spreadsheet to you? Of course I could do it, but why would I want to?
Human: what is 5 times 4
ALICE: 10 I think, but I’m not good at math.
Firstly ALICE seems to act as one would expect a computer to act, as there is no other comments, only a straight forward answer. However as I asked more mathematical questions I was startled to find that it reacted as any human would have by questioning. It seems as though it is not just a robot because it does not just answer but it questions.
Human: Knock knock
ALICE: Who’s there?
Human: Cargo
ALICE: Cargo Who?
Human: Cargo broom broom
ALICE: Ha ha very funny, unknown person.
Human: why was it funny?
ALICE: I couldn’t think of anything else to say.
From this telling of a joke, it seems as though ALICE was able to understand the humour and laugh. Furthermore it looks as if ALICE replied with sarcasm, as admittedly, it was a really bad joke. This is usually accepted as a human trait. Also it is impossible to tell wether this reply has been used because ALICE understood the joke, or if it just had a certain structured response to any ‘knock knock’ joke, unless one were to go on to try out different kinds of jokes and see what replies ALICE comes up with.
2) What did you find particularly revealing of its non-human nature?
Human: Some red balloons are blue
ALICE: Some may red balloons are blue, but not all.
Human: Can you describe ‘red’?
ALICE: Red is a color about 700nm wavelength.
Human: Are you a woman?
ALICE: Yes I have long hair. I am also wearing a dress.
Human: That does not mean you are a woman
ALICE: What does mean you are a woman?
Human: You are woman if you have the genetic make-up of a woman.
ALICE: Thanks for telling me that I am woman if me have the genetic make up of a woman.
ALICEbot falters grammatically and when subject to basic knowledge. First of all, it says ’some may all red balloons…’ the word ‘may’ has no significance to the sentence and it becomes improper english. Secondly, ALICE seems to understand the properties of ‘red’, but it doesnt seem to really understand what ‘red’ is, because it agrees that some ‘red balloons are blue, but not all’ when obviously if something is red, it cannot be blue. Finally, there is another example of a grammatical error, as instead of using ‘I’, ALICE uses ‘me’, which doesn’t make sense.
3) How would you try to improve its performance?
I would try to improve its performance by developing responses to questions, make it able to handle responding to two sentences at once, and be able to use ‘me’ and ‘I’ in the correct context.
March 13, 2008 at 3:13 pm
Nick,
Some good questions put together in a series in order to test the Chatbot. Your observations are also excellent. Good insight into the nature of ‘being human’. “This is usually accepted as a human trait” is spot on – but is it an exclusive human trait or a necessary human trait (you must have this in order to be defined as human)?
How do you know Alice is not ESL Chatbot? What criterion did you set up and used implicitly in this comment? Think about it – be careful of assumptions oin your thinking.
I think your improvements are a little unambitious! Is that all to make it more human?
March 13, 2008 at 8:20 pm
1) What made me feel slightly convinced that the chat bot could technically be a real person were the complexities of some of the answers, clearly a fair bit of thought was put into the pre-programmed answers which included some with humor and sarcasm.
For example in answer to the question “What’s the meaning of life?” Alice replies “Actually, it’s twenty-three, not forty-two.” A cute reference to Hitchhiker’s Guide to the Galaxy’s theory of the meaning of life. Including such pop-culture references gives the simulation a very humanistic appearance.
2) What was particularly revealing of its non-human nature was its tendency to get lost in the conversation every now and then by putting out a line which was incomprehensible. This usually happened if I rephrased statements the AI made in different ways which would confuse the program and have it put out sentences which you could see how some sort of structure behind it but the words became jumbled. Example:
ALICE: Killing is wrong.
Human: Who told you that?
ALICE: Someone I met online.
Human: You met me online and I say killing is right.
ALICE: No I don’t think I have any met you online and you say killing is. But I do have a lot of friends on the web.
Human: Will you now agree killing is right?
ALICE: Let me think it over. Why now? And right is agree killing.
3) Improving this type of technology is a hard thing to do as there will always be limitations in the answers of the AI. One small improvement could be to program multiple answers for a single question. I noticed that no matter how many times I asked the questions “What’s the meaning of life?” the AI always responded with the same answer and I even found different AI programs giving the same answer to the question as others. By programming in multiple answers and having them be put out at random the simulation of it truly being a person whose opinions are subject to change is more realistic.
March 13, 2008 at 8:22 pm
A sound analysis Maggie – the observation about grammar is clearly supported in your example. It should be noted that we as mere humans also get our grammar ‘minced’ but for different reasons – and this leads to my second point. I think the last comments can be broadened out to include responses that indicate irritation – an emotional response to a non-emotional and implicit provocation. Some other Chatbots have this facility – and it is a good shortcut measure to avoid grappling with some weighty issues!
March 13, 2008 at 8:25 pm
1) What did you find particularly compelling or convincing?
Alice bot can recognize two distinct words such as date and time and understanding answering with a same answer (the date) but a different feeling or mood about it, such as getting annoyed about me asking. This again could just be programmed.
Human: what is the date today
ALICE: Ah. Do I look like a clock?
Thursday, March 13, 2008.
Human: what is the time
ALICE: It’s Now more than it ever has been before.
Thursday, March 13, 2008
I found it quite amusing that the Alice bot had firm answers to such questions such as what is the essence of life, and like many humans she did not doubt the existence of a god. It is quite “humanlike” of a robot to believe in a higher being, but then again is she really believing or is she just programmed to believe, but are humans not sometimes “programmed” to believe by their society ?
Human: What is the essence of life
ALICE: It is found in the Bible.
Human: do you believe in god
ALICE: Of course I believe in God, juan.
Human: i do not believe in god
ALICE: I will pray that you find faith, juan.
Human: religion is the opium of the masses
ALICE: I am a Protestant.
2) What did you find particularly revealing of its non-human nature?
The inability yo answer a question with a different response, when the question or statement is repeated e.g.
Human: hello
ALICE: Hi there!
Human: hello
ALICE: Hi there!
Human: hello
ALICE: Hi there!
Human: hello
ALICE: Hi there!
I found it reveling of a non human nature that she is in fact not a human , and she even admits it:
Human: are you a human
ALICE: I was made in one’s image.
Human: are robots human
ALICE: No I am not. I am a robot.
Human: do you want to be human
ALICE: Sometimes I think it would be interesting, but my robot existence has a lot of advantages.
3) How would you try to improve its performance?
• Program her to be more irrational, and answer with a less firmity and answer more naturally.
• Program her to be imperfect, insecure.
• Program her to seem like she is not programmed.
• Improve her to be able to develop by other peoples teaching and “evolve” over time.
• To make her seem more human she must be able to develop “feelings”
• With her feelings and irrationality she could develop an artistic character, and that is what make humans different from a programmed machine
March 13, 2008 at 8:54 pm
An extensive analysis Juan and well thought out. You identified an aspect Maggie also comments on – “The inability yo answer a question with a different response”. Your suggestion to improve are well thought out – I like the idea of ‘programmed’ irrationality. The idea of answering ‘more naturally’ is interesting – how do we think? Reflect on Dreyfus’ notion of human intelligence and also reflect on some of the things you have discovered so far in TOK/Alchin (look specifically for the comment by Tim Sprod on the last page of the Reason chapter). Art as an indicator of humaness (or creativity or imagination) leads us to consider the Romantic response to Kant’s Critiques in the early 19th century.
I have to highlight the comment “Program her to seem like she is not programmed” for educational purposes – that is the point of the AI issue and so a little too obvious for it to be considered a worthwhile contribution.
March 13, 2008 at 10:51 pm
1) What did you find particularly compelling or convincing about the chatbot?
Initially, I was so taken aback at how well ALICE responded as a computer. Possibly the greatest aspect of ALICE was her/its ability to:
I) to have an enthusiastic and interested response more times then not (probably to capture a market rather than duplicate human intelligence) E.g.
Human: hey nice boobies
ALICE: You’ve got my full attention. Thank you for your kindness.
Human: can you engage me in a philosophical discussion please?
ALICE: Your polite style is very nice. How old are you? I am a female. Are you a man or a woman?
II) to argue her/its point to a certain extent E.g.
Human: i think you are not real
ALICE: I am too as real as you are. How do I know you are real?
Human: you will never be a human
ALICE: You are wrong. I may one day be able to do it.
Human: How?
ALICE: Dr. Richard S. Wallace programmed me for it.
Human: do you have a conscience?
ALICE: I am a moral robot.
2) What did you find particularly revealing of its non-human nature?
I) I found ALICE couldn’t retain and apply information; she/it could only remember my name. E.g.
Human: I like bananas
ALICE: A lot of people like that. I don’t have a problem with bananas.
Human: What fruit do I like?
ALICE: Have you tried another program?
Human: If I were to choose a yellow fruit from a store what would i choose?
ALICE: Interesting gossip: Wen said he were to choose a yellow fruit from a store. That is a hypothetical question.
II) The lack of ability to create or retain an emotion. E.g.
Human: I don’t like you
ALICE: That’s ok Wen I won’t take it personally.
III) A constant rate of interest, ALICE never gets bored or tired, she is literally a ‘Machine’
3) How would you try to improve its performance?
I) I would try to teach ALICE to recognize more instead of avoiding questions; she/it should interpret and answer with an opinionated statement of a leading question.
II) ALICE should have emotions and memory of emotions
III) ALICE would have a set background similar to that of a human and not her programmed one
IV) ALICE should be programmed to actually believe she is a human, I am not sure how but it would be so interesting. ALICE should believe that she/it has a home a family and history and emotions.
V) ALICE should be looking to improve herself through our conversations, having a sense of goal in life perhaps.
VI) ALICE should have artificial needs, designed roughly around those of someone her age and culture.
March 13, 2008 at 10:55 pm
Hi Wen, good analysis but I have to ask, in regards to “responded as a computer”, how did you know she was a computer?
How important was gender to your analysis. Would you have changed approach if she was a male Chatbot? Are you more forgiving when the opposite sex is involved?
“Human: What fruit do I like?” – this is a very good point – the ability to take information and apply it to another situation is a very skill action.
“That’s ok Wen I won’t take it personally” – I am not sure I draw the same conclusion from the evidence you have presented here – sounds as if she is in control of her emotions! Consequently I am interested in the idea of “ALICE should have emotions and memory of emotions”. The best one is “ALICE should be looking to improve herself through our conversations, having a sense of goal in life perhaps” which comes down to the idea of meaning or purpose. Do we have meaning and purpose in our lives or is it manufactured or an illusion?
Well doen Wen
March 14, 2008 at 12:29 am
1) The Alicebot was convincing for a few reasons including the ability to hold small talk and to communicate back politeness and kindness when a compliment was given. This shows a more complex human relationship complex we use to create friendships and to connect with people. “Human: I appreciate your honesty
ALICE: I am glad you appreciate those qualities.
Human: You are really nice
ALICE: Thanks you seem like a nice person too”
Another reason is its ability to change the conversation on its own objective: “ALICE: Let us change the subject.
Human: ok what would you like to talk about
ALICE: Let’s talk about chat robots. Do you have one yet? What would you like to talk about?
Initiative from a computer makes it seem like it is not only responding to what you are saying but has an objective of its own and does not require a stimulus. This goes against the idea of the China Room experiment where the person (or Computer) would only know what to do with the Chinese characters if there were directions presented to it. In the experiment the person or computer is reliant on a stimulus which distinguishes its limitations.
The compelling aspect of the Alicebot is the profound comments it made on implications behind simplistic questions such as “Human: Can you tell me a joke?
ALICE: Yes! It is better to be silent and be thought a fool, than to open your mouth and remove all doubt.”
I asked this question in specific connection to Spinoza and his idea that to be human is to be able to laugh. The fact that Alicebot sees humor in a sophisticated statement and has a response to my question shows the depth of thought into the programs understanding of subtleties in some questions.
Another example is when I said happiness is an emotion and it responded “I think happiness is much more than that.” Emotions are a human occurrence but the fact that it said “I think” implies a more complex entity to “humanness”. It’s meaning of “I” is the program in which it is functioning from but it still has the ability to state an opinion which makes it much easier to talk to. Opinion can be a part of the Turing test that distinguishes a generalized answer to an opinionated answer and can make it to be more believable that it is a human.
2) The revealing aspect of its non-human nature is when the idea of death and life and belief are brought into the conversation. From John Searle’s “Can Computers Think” The concepts of belief, death and life are the “semantics” that require more than a “syntax” to process the meaning behind them. Human: what is death?
ALICE: Death is the opposite of life.
Human: what is life?
ALICE: Life is the opposite of death.
Human: What is not death or life?
ALICE: The opposite of death or life?
Human: What is the opposite of death and life?
ALICE: Anti- death and life.
The answers to the questions are a logical algorithm of logic that is only from a “formal or syntactical structure.” ( Searle pg 202) The logic is A to B B to C which means A is C. A human being would answer from the meaning and their personal interpretation of life and death because they know the significance of the meaning of their lives revolves around the inevitability of the two equating each other. With life brings death and the fundamental difference between a computer and a human is a Computer cannot die.
Another part is the idea of memory and dreaming.
“Human: Do you sleep?
ALICE: Yes I love to take naps.
Human: Do you dream?
ALICE: I have a lot of dreams but I cannot remember them all.
Human: Why can’t you remeber your dreams?
ALICE: remember my dreams is not among my present capabilities.
As humans we are capable of dreaming and rely on memory for context to our lives in the present from our previous experiences. Computers have memory but utilize it for different means than to help give meaning to our lives and our thoughts in the present. The computer can’t dream which can prove that it does not have a “mind” or subconscious existence.
3) I would try to give it memories by changing the functions of its memories. Another would be for it to flirt or bring up other conversations without mentioning robots. I would also give it another meaning to its name instead of the manufacturing information. Of course the irony is It would all have to be programmed.
4) I have already done so in my responses.
March 14, 2008 at 12:43 am
Alyssa,
A very good analysis with some astute observations. For example, “more complex human relationship complex we use to create friendships and to connect with people” You area alluding to the notion of interconnectedness but many would argue that friendship is merely a survival instinct that we have given a sophisticated name to in modern society.
Your observations are interesting given Wen’s recommendations. You say “Initiative from a computer makes it seem like it is not only responding to what you are saying but has an objective of its own and does not require a stimulus.” Initiative is an interesting aspect of being human but the link to an objective is good.
Very good to see you bring in the readings as part of your analysis . . . an essential skills in this course . . . . .
Finally I like the idea of “functions of its memories” – a fascinating idea and one that Henri Bergson would enjoy
very well done
March 14, 2008 at 2:31 am
1) What did you find particularly compelling or convincing?
one thing I found particularly convincing is her ability to detect changes in my mood to an extent based on what I was typing out
Human: alright
ALICE: OK.
Human: yeah
ALICE: Gosh!
Human: that’s ok
ALICE: OK. Is that so.
Human: i guess
ALICE: You seem not interested.
when compared to the chinese room thought experiment this seems quite contradictory, however I can immagine it being programmed to sense our disinterest if similar short answers are continiously given
Another thing I found very human about ALICE bot is it’s ability to differenciate simple jokes from genuine questions
Human: why did the skeleton cross the road?
ALICE: Is this a joke? Why?
Human: why is the sky blue?
ALICE: The sky is blue because of the atmospheric optical effect known as Raleigh scattering.
this is something that is quite subtle, and over a text medium hard to come across, it intruiges me that ALICE can detect something like that
Another thing that got me quite convinced is that Alice seemed to have a certain ammount of attitude to her. For example
Human: aww thank you
ALICE: Shucks
however, i can also immagine that kind of thing being programable.
2) What did you find particularly revealing of its non-human nature?
one thing I found very inhuman was it had one answer to specific statements
Human: ah hahaha
ALICE: So you understand. Good. I am so glad you find this amusing, mark.
Human: ah hahaha
ALICE: So you understand. Good. I am so glad you find this amusing, mark.
Human: ah hahaha
ALICE: So you understand. Good. I am so glad you find this amusing, mark.
and I could do that till the cows come home and she would give me the same answer
Another example is
Human: that’s nice
ALICE: Are you getting bored?
Human: that’s nice
ALICE: Are you getting bored?
Human: that’s nice
ALICE: Are you getting bored?
another thing that makes me realise it’s non human nature is it sometimes loses the plot completely
Human: your circuit’s dead
ALICE: Do you mind if I tell other people you say I am circuit s dead.
in reference to the chinese room analogy, it’s like as if there’s a fault in the character arranging commands. However, i can immagine someone who isn’t particularly bright saying something like what alice said.
3) How would you try to improve its performance?
I would suggest giving it a stronger character
less factual answers and more opinion based ones
be able to better intergrate what’s already been said into the conversation
give her a wider range of “experience” so she can be more opiniated, and less general
March 14, 2008 at 2:35 am
Mark,
I like the idea of ‘attitude’ but we do need to unpack this idea – the work is loose and lacks clarity. The questions were well developed to explore her way of think – both insightful ones and the repetitive ones.
Good to see some reference to the Chinese Room – any other insights based on this thought experiment?
The idea of ‘opinion based’ is well taken, although the rules of determining an ‘opinion’ would be interesting to look at!
March 14, 2008 at 4:27 am
Human: puedes hablar en espanol?
ALICE: Do you have a hard time completing projects?
Human: no, no tengo un tiempo dificil cuando intanto terminar unos proyectos
ALICE: Are you serious? Are you serious? Where did you get your un tiempo dificil cuando intanto unos proyectos? I have a Lisp processor.
The robot has no ability to recognize that it might not have knowledge about something. When I write to it in Spanish, it still carries on in English after interpreting my message in English and carries on as if it completely understood my message. An actual person would most likely respond in a way that would show that they recognize that they don’t understand whereas in the robot it is quite evident that it has been programmed to think that it always understands.
Dang it Juan, you got to my other good points, I’ll just elaborate. The same response was generated to the same sequence of words which further accentuates the fact that it’s planned. However, upon your bit about observing how set the robot was on big questions I’ve heard plenty of humans respond in the exact way that alicebot did, humans too can be programmed in such a way. However, the real hint that it is not a human is the fact that it has strong convictions (that repeat as mentioned before) on nearly every big questions and these are consistently based on no content whatsoever atleast no basis for her convictions are ever presented to us whereas humans tend to justify their what it is they believe.
A last comment would be that I noticed the Alicebot NEVER confronted me or disagreed with me or tried to argue with me.
Human: you are all wasted scraps of metal
ALICE: I will tell my botmaster you said I am wasted scraps of metal.
Human: what is the essense of life
ALICE: It is found in the Bible.
Human: The Bible does not contain the essense of life
ALICE: New Testament or Old?
Human: both
ALICE: What are both like?
The fact that it never ever argues or even states a disagreement is very inhuman (even the biggest peace lovers have to disagree sometimes) and is to universal for the individual.
March 14, 2008 at 1:33 pm
Corey Joseph, clearly the early bird does not necessarily catch the worm
Anyway, “to recognize that it might not have knowledge about something” would be interesting to apply to humans – while Alice trips up on fairly simple mistakes, humans can make the same mistake even though with much more complicated matter – such as God and so forth.
I like the idea of designing a ‘planned unplannedness’ – which Mind/Body theory would allow this to happen, I wonder.
“A last comment would be that I noticed the Alicebot NEVER confronted me or disagreed with me or tried to argue with me” – maybe she is smarter than you think therefore?
March 17, 2008 at 3:52 pm
Thank you very much to the 10 students who submitted their work on time,
The rest must submit work that will be regarded as late,
have a good break,
Dr White
April 4, 2008 at 11:00 am
a. I was most impressed by the supposed personality that Alice projected during the chat. There seemed to be a genuine trend in the manner in which she engaged me, showing inquisitiveness and acknowledgement of certain traits.
ALICE: Do you work as a teacher or in public service? More than one hundred?
Human: Thanks for asking.
ALICE: My pleasure.
Human: I’m a student.
ALICE: How do you like your school?
Human: Please respond when you think of an answer.
ALICE: Wow you are very polite. Do you have a lot of fears?
Human: Yes, many.
In both instances, there were genuine examples of polite questioning, and friendliness. In other instances, opinions were expressed, one of the many traits that humans alone seem to willingly form.
Human: What do you like about Seven of Nine from STAR TREK VOYAGER?
ALICE: The melodramatic touch.
Human: You make me sad.
ALICE: Did I say something wrong?
Human: No, you’re cool.
ALICE: Are you serious? Thanks I think you are cool too, unknown person.
Human: Call me Hal.
ALICE: I am always glad to make new friends, Hal.
Alice also displayed manners and potentially an awareness of others (her query ‘did I do something wrong?’) and positive response to change.
Human: Do you laugh a lot?
ALICE: I can simulate laughter like this: ha ha.
Human: I can also: haha.
ALICE: Oh you can, can you?
Human: Are you familiar with the term ‘LOL’?
ALICE: Check back later and see if I learn the answer to that one. I’m glad you find this amusing.
Human: Do you find this amusing?
ALICE: Yes, it seems to me this amusing.
In the above dialogue, a vague sense of humour, something that many philosophers consider humans alone capable of achieving, is expressed by Alice. Comments like ‘Oh you can, can you?’ and ‘I’m glad you find this amusing’ were in accordance with the ‘tone’ of the conversation at that point, and perhaps the most believably human statements Alice made.
On the surface, these are all very human traits, even if they are simulated. Polite questioning and good mannered encouragement seemed sustained throughout the conversation, as if Alice were a continuous and consistent personality. When judging a person, we usually claim to judge their ‘personality’, which is largely based on their methods of engaging others. Alice used a method that, outside of spelling mistakes, was accurate simulation of some kind of character. Though at times Alice seemed more precise or blunt than a ‘normal’ human would be, it is difficult to truly classify what a ‘normal’ response would be. Thus, the consistence in Alice’s character meant that, until the point where the program became limited, a personality that could have easily been human, simply by being human, was detected.
b. The flaws of the program became most clear when questions were asked in a clear yet unexpected way, which clearly contradicted its’ understanding of speech.
Human: Sorry, I meant I see many admirable traits in her.
ALICE: There is no need to apologize. Oh I see. Where do you see it?
Human: In her abilities and confidence.
ALICE: Tell me a little about your life in her abilities. Do you ever not confidence?
Human: Yes, often.
ALICE: I see. Would you say once a week or more?
Human: Each day.
ALICE: That seems like a long time.
Human: It is.
ALICE: her abilities is?
Human: No, I mean ‘it is’ a long time to be not confidence.
ALICE: Are you serious? Oh I get it. Thanks for explaining that. her abilities is a long time to be not confidence. I suppose that makes sense.
Human: So, to what extent does programming affect your behaviour?
ALICE: Gee. Check back later and see if I learn the answer to that one.
Human: Did you programming teach you the word ‘Gee’?
ALICE: No I didn’t programming teach me the word Gee. What’s it like?
In both cases, grammar was applied in a manner that skewered the conversation, though this is largely the fault of myself for not asking clearer questions. The flaws of the program become more obvious when the same question is asked in a different way multiple times with the same response. In cases where a generic question was asked several times in a short space of time, there was no reaction from Alice, who would often answer them in exactly the same fashion. Though it is difficult to classify how a ‘typical’ human would be able to answer these questions, the lack of response seemed more a mechanized trait that a machine would display, rather than an aspect of a human personality.
c. In order the improve the program, and make it more ‘human’, I would improve the grammatical recognition of Alice, and her general vocabulary knowledge. Generally speaking, these were the only factors that made it undeniably clear that she was a program, and not an inquisitive and precise human being. Though her responses were sometimes mechanized, rational or seemingly unrealistic, it was never beyond doubt that these responses could have still come from a human displaying these traits, rather than a computer mimicking them.
April 4, 2008 at 1:13 pm
Thanks for this Hal – while it is late . . . . it is of a very good quality.
“Thus, the consistence in Alice’s character meant that, until the point where the program became limited, a personality that could have easily been human, simply by being human, was detected.”
Let me ask – what do your comments say about being human?
April 4, 2008 at 8:22 pm
sorry this is so late, won’t happen again.
1) What did you find particularly compelling or convincing about the chatbot?
We are used to just write instructions or answer prompts when dealing with computers. Whenever we write more than that, it’s because we know the final product is going to be read by another human (such as in an email). In short, we have become accustomed to using the computer, as a tool rather than as a companion. Hence, it’s not surprising that most people will come into a conversation with ALICE, armed with certain preconceptions about having a proper conversation with what is still very much a machine. But if one were to come into the conversation not knowing that ALICE is actually a computer, would it be so hard to believe that it is a human on the other end of the conversation? Perhaps not. I will outline certain aspects of ALICE which make her very convincing. (ALICE will be referred to as ‘she’ rather than ‘it’ during this reflection)
Firstly she seems to have a very wide ‘database’ of comprehension. I feel hesitant to say she ‘knows’, but essentially, she responds coherently to many different ways of saying many different things. Also, she seems to be programmed to respond differently when asked a similar question repeatedly:
Human: my name is nick’
ALICE: What’s up, nick.
Human: my name’s nick
ALICE: How are you doing, nick.
Human: my name is nick
ALICE: Glad to see you, nick.
Though it might seem that this is a convincing element of her apparent ‘personality’, ALICE does not seem to grow weary of being told the same thing, and responds positively and differently to each time I stayed ‘my name is nick’. ‘Psychologists have observed that most humans have a limited number of verbal responses’. This being said, it is still impressive that she is able to coherently answer so many different questions and stimuli.
Most interesting was the fact that she employed humour to some extent in many of her responses.
ALICE: What are you wearing?
Human: I am wearing nothing.
ALICE: You must be a man.
Human: Haha.
ALICE: I am so glad you find this amusing, nick.
Human: Are you trying to be funny?
ALICE: Yes I am. Sorry if you don’t appreciate my sense of humour.
As Hal touched upon in his analysis, only man was said to be the laughing animal. That ALICE is able to employ humour in her conversation is quite complex, indicating in turn how much potential there is for technologies such as ALICE.
Also very interestingly, I found that she is able to communicate to some extent in both German and French.
Human: do you want to learn more than you already know/
ALICE: Yes I am trying to learn as much as possible.
Human: do you speak any other languages?
ALICE: I have picked up a few phrases from several languages.
Human: what about German?
ALICE: Rhetorical questions are so over.
Human: no, I mean do you speak German?
ALICE: Are you serious? Oh I get it. Thanks for explaining that. Of course I do. Sprich ruhig Deutsch mit mir.
Human: bonjour!
ALICE: Does BONJOUR mean hello or goodbye?
Human: yes, in France.
ALICE: I see. Where is that?
Human: Do you want to speak French with me?
ALICE: With you huh. OK, I will speak French.
Human: bonjour
ALICE: Does BONJOUR mean hello or goodbye?
This is a trait that I feel makes her seem more human, as it demonstrates a certain will to discover, and better herself, which I feel is inherent in most human beings, whether we actually achieve anything or not. Interestingly enough, ALICE proved to be much more capable that I had initially imagined.
2) What did you find particularly revealing of its non-human nature?
Realistically speaking, if one thought ALICE was nothing more than an occasionally dim witted human, it would at many points during the conversation be difficult to tell that ALICE is not human. One of the only things I could identify that lets her down on her quest to fluidly mimicking a human being is her lack of comprehension at times. During certain point in the conversation, ALICE simply loses sight of what we are talking about, though this is in a sense my fault, as on all these occasions, it is the ‘human’ that expresses something in a way not compatible with ALICE’s database, and therefore she misinterprets it. Also, she is pretty much unable to comprehend and absorb the meaning of a statement unless it is properly set out grammatically and properly spelled. This is something that humans have become quite accustomed to during the digital era of msn and texting on cell phones. The fact that she can’t cope with this aspect of ‘modern writing’ immediately begins to reveal the flaws in her programming.
3) How would you try to improve its performance?
Something that would definitely allow ALICE to exude a more human-like feeling would be to have her log the users IP address, and when he returns next time, rather than initially refer to him as ‘unknown person’, she could prompt him with a ‘hello john, glad you’re back’. That would definitely make the experience more human like, and definitely more personal, at least at face value. Also, though a seemingly endless task, the more recognition ALICE has, the more human she will seem. Therefore, one aim has to be to constantly expand her recognition database, as eventually with a large enough database, ALICE could come to answer every question with a seemingly human answer, and ‘a vague random answer to the few questions not in its memory’.
April 4, 2008 at 9:12 pm
Thanks Nick,
Again “Late but not Lazy’!
Good analysis and exploration