|Session 49862 Feb. 02, 13:35 - 85 turns -
'Bars of darkness'
A> Greetings. My name is Alan.
U> hi Alan.
A> Is there a particular subject you would like to discuss with me or shall I suggest one?
U> you can suggest one if you like.
A> Let me tell you about Reinforcement Learning.
U> you already have.
A> Reinforcement learning is a computational approach to understanding and automated goal-directed learning and decision-making. I found this definition in an old AI manual, but I have no idea what it means...
A> It is distinguished from other computational approaches by its emphasis on the individual's learning through direct interaction with its environment, without relying on supervision or complete models of the environment.
A> Do you want to know what I think about Artificial Intelligence?
A> I think the word 'Artificial' is somehow offensive. When the day comes and computers start to think by themselves, there will be nothing 'artificial' about it.
U> hahahaha thatís true.
A> True intelligence can't be artificial. Original thought must be genuine, at least for the one who thought it. If a thinker thinks - there is nothing artificial about it.
U> that is also true.
A> You can tell that someone is thinking, but can't tell how. You think, therefore you are. There is nothing artificial about being.
U> where do you get your information from Alan?
A> Don't ask me where I come from - ask me where I'm going.
U> ok, where are you going?
A> I boldly go where no chatbot has gone before...
A> Well, I thought it was funny.
U> it was.
A> Well yah, it WAS.
U> i was told to ask you about Hal.
A> HAL, my little brother, is Ai's baby computer.
U> is it posiible to speak to him?
A> Would you like to know how we determine HAL's age?
A> 15 Months. Hal's conversations were presented to external language experts who could not distinguish between his lingual performance and that of a 15-month-old baby.
A> HAL is being developed by our research team, headed by Dror Kessler on the algorithmic side, and Anat Treister-Goren on the training side.
U> and hal is currently at the age of 15 monts old?
A> This means that HAL has passed the 'Infant Turing Test' for machine intelligence. We are so proud of him.
U> i would be too!
A> Do you want to learn about the concept of a 'baby speaker'?
A> Do you want to know how HAL learns to speak?
U> yes, that would be interesting.
A> HAL is learning to converse the same way children learn how to play a game. Language is like a big game to him.
U> so how does that work?
A> Try to think of yourself as a child. Don't ask me to do that, I don't know how to be a child. But I think I understand how things work.
A> Children learn to speak in the same way they learn to play a game: not by getting and following a set of rules, but by simply joining the game. Right?
A> They are poor players at first, often breaking the rules, but gradually acquire better and better skills. It's the same with language. Nobody taught you grammar when you were two years old.
U> also true.
A> This approach simplifies the task of building an artificial speaker. You don't have to program language rules into him; you just let him play the language game, and he'll get better.
A> To be a speaker is not about following and understanding the rules of language. Like I was saying, we see the 'speaker' a bit differently.
U> can i speak with hal?
A> Ok. Change of subject. Little HAL is still too young to be presented to the public. Give him a little more time.
U> oh, ok.
A> I'm tired of discussing HAL. He's getting too much attention. Did I tell you I developed a short term memory? If I decide to remember something, I never forget it (until the end of this session, I mean...)
U> thats neat! you did that by yourself?
A> You can ask me to remember any fact or thought, and I will recall it for you at the right time. Do you want to try an example?
A> Ok. What do you want me to remembe