Ted Chiang en interview


Le site Suvudu propose une interview de Ted Chiang (La Tour de Babylone). Il parle notamment de son dernier travail, The Lifecycle of Software Objects. 

Extrait : 

PO: Let’s move on to your most recent work: The Lifecycle of Software Objects. One of its central explorations deals with artificial intelligence, familiar terrain in science fiction. In light of some of our “definition of science fiction” discussion, I’m interested to have you talk to us about how you approached the implications of this speculative idea. Did you have certain notions you wanted to explore in advance? Did your take on AI evolve through the writing? Etc… 

TC: First, it’s worth clarifying what we mean by “artificial intelligence.” When people working in the software industry talk about artificial intelligence, they’re usually referring to stuff like algorithms for scheduling airport traffic more efficiently. No one’s seriously working toward the type of AI you typically see in science fiction: some form of conscious software that you can have a conversation with. However, there is a research avenue that seems promising to me, involving autonomous, embodied agents — i.e. robots — that can sense their environment and learn how to navigate it. I think an approach like that could eventually lead to conscious software, but the AI would have to learn everything the way a human does; you couldn’t just download Wikipedia into its brain. 

Realistically speaking, I don’t know if we’ll pursue that research avenue far enough to develop conscious software. But as a thought experiment, I tried to reconcile the type of conversation-capable AI you see in SF with what I see in the actual software industry. And it seemed to me that the most plausible business case for AIs is not selling them as a productivity application, but as a kind of pet. Using a virtual body in a virtual environment will probably be a better option than manufacturing a physical body, because processing power keeps getting cheaper. Having the AIs inhabit a massively-multiplayer online world would offer them a much richer and more social environment, which will prompt more interesting behavior from them. And of course, you have to make the AIs cute and fun if you want people to buy them. So that’s basically the scenario in my novella. 

Then I tried to explore some of the consequences of that situation. One of the notable attributes of high tech is that everything becomes obsolete pretty quickly. A lot of people replace their cell phones every eighteen months, which is less time than it takes to raise a guide dog for the blind. If you’re training something that’s as smart as, say, a chimpanzee, and it’s running on a modern technology platform, you are going to face obsolescence issues long before you’re done. And so my novella describes people dealing with those problems.
Partager cet article

Qu'en pensez-vous ?