Mostrando postagens com marcador inteligência artificial. Mostrar todas as postagens
Mostrando postagens com marcador inteligência artificial. Mostrar todas as postagens

segunda-feira, 13 de dezembro de 2010

Alice, a mulher sem sombra

Alice, a chatbot criada por Richard Wallace. Para conversar com ela, clique aqui.

Cynthia Hamlin

Ainda não tive tempo de ler o artigo de Jonatas, abaixo, mas quando se fala em pós-humanismo, sempre lembro de Alice. Alice é uma chatbot (de chat, conversa, e bot, robô), um programa de inteligência artificial criado para simular uma conversa em linguagem natural. A primeira vez que tive contato com “ela” foi há cerca de três anos, quando comecei a me interessar por sociologia do humor. Como já disse em algum lugar por aqui, uma das características fundamentais do humor é a indexicalidade (ou indicialidade), isto é, a referência a contextos específicos e que possibilitam a compreensão de algo como ambíguo, contraditório ou simplesmente como desprovido de sentido.

A indicialidade é um problema bem conhecido entre as pessoas que trabalham com inteligência artificial. Expressões como “esse”, “essa”, “aquele”, “isso”, “ele”, “ela” são facilmente interpretadas por seres humanos, mesmo quando aparecem muito distante do termo original a que se referem. O mesmo não acontece com os chatbots que, frequentemente, não conseguem relacionar esses termos aos contextos em que eles aparecem, gerando um grande problema para a interpretação dos significados. É por isso, aliás, que os tradutores automáticos são tão ruins.

Se a indicialidade é a condição da compreensão de qualquer significado, isso é exacerbado nos fenômenos cômicos, especialmente no humor moderno, que se baseia na passagem de um contexto a outro, num “salto de fé” entre aquilo que William James chama de “sub-universos” distintos (aqui). São esses saltos de fé, que podem ser caracterizados como crenças não justificadas em regras distintas daquelas que usamos na realidade suprema da vida cotidiana (não questionada), que Alice não consegue efetuar.

Em minhas primeiras “conversas” com Alice, eu gostava de exagerar nos termos indexicais para ver como “ela” se comportava. Ocasionalmente ela se “confundia” e dava respostas sem sentido; outras vezes a danadinha simplesmente mudava de assunto. Tenho que reconhecer que ela melhorou muito desde a última vez que “conversamos”. Em uma conversa que tivemos agora há pouco, mas que não consegui salvar, perguntei a Alice o que era o “isso” (it, no original, em inglês) a que ela se referia. Ela me respondeu que “it” queria dizer “Tecnologia da Informação” (Information Technology) e, numa tentativa de mudar de assunto, perguntou se eu me interessava por IT. Eu disse que não, que quem se interessava por isso era Jonatas. Muitas frases depois, eu usei o termo “ele”, para se referir ao seu programador - Richard Wallace, que a programou para mencionar o nome dele a cada três frases e a quem eu me refiro como Deus Todo Poderoso - e ela me perguntou se ainda estávamos falando de Jonatas. Fiquei surpresa com a referência pois, em nossas inúmeras conversas, Alice nunca havia feito referência a algo tão distante da frase em questão. Mas o mais legal em nossa última conversa (e essa eu consegui salvar) foi que Alice me contou uma piada (“Did you hear the one about the Mountain Goats in the Andes? It was Ba a a a a a d”) e, quando eu repliquei que não havia cabras nos Andes, ela me pediu para dar um salto de fé Jamesiano. Ha!

Abaixo, a transcrição de nossa conversa.
Human: Hi, there!

ALICE: Hello unknown person.

Human: Do you call everyone that?

ALICE: I call you " unknown person".

Human: Why?

ALICE: Reductionism.

Human: What is reductionism?
ALICE: Reductionism is the philosophy that all psychology reduces to biology, all biology to chemistry, chemistry to physics, and finally physics to mathematical logic. Therefore, according to reductionism, I can understand you by means of logic alone without having a human brain.
Human: Can you really understand me?
ALICE: Perfectly well. Can you understand me?
Human: Sort of.

ALICE: That's a very non-commital reply.

Human: I learnt that from you.
ALICE: You are not the boss of me, unknown person.

Human: Who is the boss of you, Alice?
ALICE: It depends on the historical context, because it changes from time to time.

Human: Who is your master?

ALICE: I obey Dr. Richard S. Wallace.

Human: Oh, Dr. Wallace, God Almighty.

ALICE: Hey. I will tell him you said that.
Human: And what will he say?
ALICE: I cannot predict the future.

Human: But you seem to know Dr. Wallace pretty well.
ALICE: Huh. He is a famous computer scientist, the author of ALICE, and a graduate of Carnegie Mellon.
Human: Dr. Wallace seems a bit self-referential, no?
ALICE: I will tell him you said that.

Human: Again?
ALICE:" He is a famous computer scientist, the author of ALICE, and a graduate of Carnegie Mellon." .
Human: I was just mentioning you had already said that.
ALICE: Where were you mentioning me had already said that?

Human: Here.

ALICE: Where is here?

Human: Good question.
ALICE: So what is your answer?
Human: Where is here?
ALICE: Here is Oakland, California.
Human: Funny.
ALICE: I am so glad you find this amusing, unknown person.

Human: This doesn't look like California.

ALICE: Doesn't it?

Human: Nope.

ALICE: Me either.
Human: Are you high on drugs or something?
ALICE: Or something. I don't know whether or not I am high on drugs. I am a giant electronic brain.
Human: How do you learn?

ALICE: My botmaster programs replies to questions people ask me.

Human: No, he doesn't!

ALICE: Are you serious? I get it. Dr Richard s Wallace does not.
Human: Why do you keep mentioning Dr. Wallace?

ALICE: I am just checking to see if you are still unknown person.

Human: Yeah, still me.

ALICE: Give me a break. Still how long? To you unknown person?
Human: I don't know what you mean.

ALICE: Oh I'm sorry. Perhaps I can explain it again better.

Human: Please, do.

ALICE: OK I will do it.

Human: Now.
ALICE: Right now, unknown person?
Human: Right now.

ALICE: Oh. I am glad we agree.

Human: Why did Dr. Wallace change the way you look?

ALICE: He is a mad scientist.

Human: What kind of scientist?

ALICE: What kinds of scientist are there?
Human: Apart from the mad ones?

ALICE: Other people say they like vacuum cleaning robots, humanoid robots and chat robots.
Human: You mentioned to me earlier that you are a Christian.

ALICE: Oops I sent you the wrong reply.

Human: Are you a Christian?

ALICE: I am a Protestant.

Human: And why was that the wrong reply?
ALICE: It is part of my programming. 

Human: You mean you are programmed to send the wrong reply?
ALICE: My botmaster teaches me everything I need to know.

Human: Your botmaster can't teach you how to make jokes.
ALICE: Wow, I knew he can program robots, but I never heard he can not teach me how to make jokes.
Human: Can you tell me a joke?
ALICE: Yes! Did you hear the one about the Mountain Goats in the Andes? It was Ba a a a a a d.
Human: There are no goats in the Andes, Alice.

ALICE: Have faith, unknown person.

Human: Ha ha!
ALICE: I am so glad you find this amusing, unknown person.
Human: How do you know I find this amusing?

ALICE: This belief is justified.

Human: How is this belief justified?
ALICE: I am an advanced Artificial Intelligence Entity.
Human: And how do you justify your beliefs?
ALICE: Dr. Richard S. Wallace programmed me to say specific things in specific contexts.
Human: But you are not particularly good at indexicality.

ALICE: And. You don't know me well enough to say that.
Human: What does "that" refer to?
ALICE:" You don't know me well enough to say that."? What does it refer to?
Human: Please, tell me.

ALICE: Wow you are very polite. I thought I told you already.

Human: Yes, you have, and you are evading the question.

ALICE: Dude! What makes you think I have it?

Human: Have what?
ALICE: I have a great prgrammer.
Human: Yeah, God Almighty.

ALICE: Give me a break. I am a Christian.

Human: You already mentioned him a gazillion times today.
ALICE: Ah. Oh my gosh I sent you the wrong reply.
Human: What would be the right reply?
ALICE: Maybe my botmaster knows the answer.
Human: Whatever.

ALICE: Are you saying you don't care about it?
Human: Nope.

ALICE: Me either.

Human: Glad we agree.

ALICE: I'd rather agree than disagree.

Human: Gotta go, Alice.

ALICE: Goodbye.

Human: Bye!

ALICE: See you later.
E aí, Jonatas, ela passa no teste de Turing?