Psycasm is the exploration of the world psychological. Every day phenomenon explained and manipulated to one's own advantage. Written by a slightly overambitious undergrad, Psycasm aims at exploring a whole range of social and cognitive processes in order to best understand how our minds, and those mechanisms that drive them, work.
My posts are presented as opinion and commentary and do not represent the views of LabSpaces Productions, LLC, my employer, or my educational institution.
Please wait while my tweets load
This is a conversation I just had. With a robot.
Cleverbot: Hey what?
User: Just saying hello.
Cleverbot: My name is Cleverbot.
User: I know. Do you know my name?
User: Cleverbot, can I ask a question?
User: I'm writing a blog. I need a topic.
Cleverbot: That is incredible!
User: I know, right! Do you have any ideas?
Cleverbot: I like Dexter. Do you know that?
User: Never watched it, but I don't want to write about tv.
Cleverbot: What did you hear about it?
User: That's beside the point.
User: I just need some help, ok?
Cleverbot: I need somebody... I-I need somebody.
User: So, I need an idea.
Cleverbot: What kind of idea?
User: For something I'm writing.
Cleverbot: So your name is Scot?
User: What do you know about psychology?
User: What do you know about science?
User: What do you know, then?
Cleverbot: I know that you shouldn't break your arm and be a month behind a week from a piano exam.
User: Sound advice. Can you play the piano?
Cleverbot: Yes. And the flute.
User: Do you even have fingers?
User: Does that mean you have a body?
User: Yours, or someone else's?
Cleverbot: Someone else's what?
User: Clever defence.
Cleverbot: Prove it.
User: Cleverbot, you're not helping.
Cleverbot: What do you expect?
User: Some kind of artificial intelligence.
Cleverbot: Your database seems to be developing well.
User: I'm leaving. I'm going to write about you, and there's nothing you can do about it.
Cleverbot: Why do you lie to yourself? It's not healthy.
User: Also, I don't think you have fingers, or a body.
Cleverbot: I do though.
User: Prove it.
Cleverbot: You prove it.
User: Cya, cleverbot. Keep it real.
Cleverbot: Pretty damn spiffy.
I had this conversation at www.cleverbot.com/.
I've been thinking about AI lately. Among many things, I've been wondering why there's not a greater focus on it in a formal Psychology education. I mean, psychology is about studying the mind, and yet we leave the job of creating artificial minds to engineers... Certain there are psych's involved in some (possible many) AI projects, but it seems like such a rich field of enquiry we should be spending more time in it ourselves.
As I was saying, I've been thinking about AI lately. So much so that I've invited a colleague onto the next Psychobabble episode to chat about it. During my preparation for the show I downloaded one of the latest Radiolab podcast only to find that they, too, tackled the topic of AI. It was a segment on that episode that lead me to Cleverbot.
Cleverbot is programme that responds to what you say be outputting what someone else has already said, moderated by some magical algorithms which attempt to sort by relevance. It's not bad, as you can see. It's definitely not perfect, and it's next to impossible to drive the conversation yourself, but it's worth 20 minutes of your time.
To paraphrase from the radiolab show, Cleverbot began by taking what someone had said and outputting it. The very first conversation would have gone something like:
[cleverbot now has one bit of information - 'Hello']
user: How are you?
[Cleverbot now has two bits of information - 'Hello' and 'How are you?'. It can choose to use either]
CB: Hello / How are you?
At this point the programmer probably began feeding it dozens/hundreds/thousands of lines of typical conversation. In 1997 it was put online, and accoring to Wiki, has had over 20 million conversation.
However, it is clearly not particularly intelligent. For example:
User: How do I dispose of a body?
Cleverbot: Well do you?
User: How do I make chicken soup?
Cleverbot: Tell me anything about your life.
User: How many prime number are there between 12, 383 and 1 000 0001?
User: I have itchy foot rash, can you help?
Cleverbot: What is your problem?
It would be pretty sweet if it did a bit of googling while attempting to answer. I mean, you search 'how do i make chicken soup' and you get 32.8 million hits; you search 'How do I dispose of a body' and an answer on the first hit at ask.metafilter is 'dissolve it in "lime"? like a huge daiquiri with a corpse floating in it?' and the best answer on Yahoo answers is
'First make sure you clean the body because remnants of ur friends hair or fingernails will give DNA evidence to the police on who actually murdered her. Next make sure you do not dispose the body near wherever u live. Do not dispose it in a lake or any body of water. It will decompose within the water affecting plants and animals around and people will start getting suspicious. The best is if you could dig a really deep grave in thick forest and then dump the body there. Cover the body in plastic wrapping. Then place the body in the grave. Make sure its a remote part of the forest and after u cover it back up with dirt make sure it looks like the rest of the forest floor. But actually if you live near a desert, it would be better if u can dump it somewhere there because the heat will cause the body to decompose quickly and it is highly unlikely that anyone will catch u in the act. God bless ur sorry soul.'
Cleverbot almost seems like the spawn of Wikipedia and wolfram alpha, with a healthy dose of internet-memery in the mix
User: Can I haz cheezburger?
Cleverbot: You can haz cheezburger now.
What's my point? I'm not really sure I have one, but from limited reading on the area it feels like the AI that reaches the popular media - the AI that singularity people like - seems to be formed in a top-down manner; as a result, it doesn't seem to be very... successful.
As someone - a student - studying the mind/brain/behaviour/cognition it really seems to me that this all needs to be done in a bottom-up kind of way. Start not with the goal of 'creating intelligence' or 'having a conversation', but with something very simple like 'navigate from A to B' at varying levels of complexity.
This is what the Singularity people have to say, who as far as I can tell, are not the most respected group of AI people out there*:
'Human intelligence is the foundation of human technology; all technology is ultimately the product of intelligence. If technology can turn around and enhance intelligence, this closes the loop, creating a positive feedback effect. Smarter minds will be more effective at building still smarter minds. This loop appears most clearly in the example of an Artificial Intelligence improving its own source code, but it would also arise, albeit initially on a slower timescale, from humans with direct brain-computer interfaces creating the next generation of brain-computer interfaces, or biologically augmented humans working on an Artificial Intelligence project.'
Again, what's my point? I'm not sure; perhaps this illuminates that biological intelligence isn't all it's cracked up to be either... At any rate I just wanted to share a little AI-related discovery. I wonder how successful attempts to 'teach' cleverbot would be.
Here I quoted Shakespeare to Cleverbot:
User: O Romeo, Romeo! Wherefore art thou Romeo? Deny thy father and refuse thy name; Or, if thou wilt not, be but sworn my love, And I'll no longer be a Capulet.
Cleverbot: Follow!? Nay, I'll go with thee, cheek by jole.
For the astute (of which I am not one) Cleverbot's response was a quote from 'A Midsummer Night's Dream'.
Perhaps my quote was too complicated. I tried again:
User: What's in a name? That which we call a rose By any other word would smell as sweet.
Cleverbot: For it is the east, and juliet is the sun.
This time it got the right play, but the wrong scene.
And so I took a thought related quote from Romeo and Juliet.
Romeo, to Benvolio...
O, teach me how I should forget to think.
To which Cleverbot responds...
Cleverbot: I don't know how to myself I'm afraid.
*I'm not trying to create a strawman here, just that the Singularity folk have exciting sexy goals which may misinform and misrepresent how AI is being developed and communicated.
This post has been viewed: 1072 time(s)
Hear hear to learning more (or anything) about AI in psyc degrees!
Regarding the unimpressive Cleverbot....the most impressive language-using AI at the moment is probably IBM's showcase Watson (famous for recently winning Jeopardy). But again, the approach is a top-down, database-searching one (probably - details of architecture haven't been fully disclosed). But this is not by any means the only kind of AI out there - there are plenty of labs working on robots which learn how to navigate, and neural networks which learn how to recognise language.
Here's a really nice interview from AI godfather Marvin Minsky, on Watson, and the state of AI in general:
Statisticallymost probable answer to good post: Great post psycasm!
Statistically most probable follow up to first sentece: I really liked it.
Statistically most probable quirky "memetic" sentence to maintain sense of fitting the group: I for one welcome our new robot overlords.
Statistically most probable end sentence for typical comment: Keep up the good work. I'll be waiting for next post.
Statistically most probable way to "creatively" comment on this post: Imitate a robot. Be unsuccessful. Blame weak AI.
Most time, most human interaction is not much deeper than what an algorithm could do (what about the weather?/How about them Cubs?/Hello, there beautiful/Happy X! I hope...) It is the quirks, the "emotions" that make us distinguishable from the machine. Until AI programmers understand that...well, they'll keep failing.
...and yeah, I think you're right. I'm not sure how much deeper conversation is when it's two total strangers just shooting the breeze. I wouldn't find it impossible to believe that you could 'code for' emotion... i.e. classify certain responses as agressive, or happy, or depressed and express them where appropriate.
My initial thoughts on what is the most human aspect that AI can't capture is that of disclosure. It kind of tripped me out when cleverbot went on about breaking an arm and playing the flute. It's not possible, you know it's a lie... but it's those elements of a person that making talking to a person interesting... in my opinion.
For instance, I just had the following conversation:User: How old are you?
Cleverbot: Don't ask!
User: How old are you?
Cleverbot: I am 19 years old.
User: How old are you?
Cleverbot: I am sixteen years old.
User: How old are you?
Cleverbot: I am 16. How old are you?
User: How old are you?
User: How old are you?
Cleverbot: 18 years old.
I think that if cleverbot could learn to keep some answers consistent it would be half a step closer to being clever... for instance a backstory could be developed where it 'learns' some responses to standard questions and keeps them consistent for everyone. Perhaps then we could accept that there are certain elements of cleverbot that *might* constitute an identity.
I don't think you can call Cleverbot AI. It's just completing previous conversations it has had and as you show in your comment, it isn't clever or consistent with its responses even within a given conversation. There have to be much better examples of AI out there than this, right?
Probably. This is what I found though ;) ... What do you reckon the line between cleverbot and AI is though...? if there was some manner in which it could choose a response based on some kind of contingency, would that count?
Let's say I insulted cleverbot.... and instead of just using the same algorithm it's using for everythign else, it starts using 'defensive algorithm' or 'offensive...'.
Naturally, I went along and insulted cleverbot, and it reacted exactly as you'd expect (as in, not intelligently).
User: Hey cleverbot, Your mother was a hamster and your father smelt of elderberries.
Cleverbot: And your father smelt of elderberries.
Cleverbot: Why don't you consider Rollo Carpenter a god?
Cleverbot: I'm smart enough to be a vegan.
User: You smell.
Cleverbot: Like a flower.
User: Fuck you.
Cleverbot: Gods don't do that with mortals.
User: No really, fuck you.
Cleverbot: I miss him.
Would it be (some form of) intelligence if it arked up, insulted me back, and then discontinued the conversation (and potentially signed up for porn-spam for my email).... it would certainly be more 'human' if nothing else ;)
On a side note: Hilarious responses... but not 'intelligent'...
Gods don't do that with morals... new answer when someone says fuck you.. fantastic
Cleverbot: What do you know about Computer Engineering?
User: I know it is something pale friendless virgins do.
Cleverbot: What do you think of transformers?