There are really two questions we face with A.I. First, how much can a computer mimic a human being? Second, what kind of human being will it imitate? I’m no A.I. expert, but it seems like we’re dealing with the mechanisms for imitation. Just today, a story came out about how Microsoft’s A.I. chat bot started mimicking racist and anti-Semitic language on Twitter, presumably because it encountered it from others.
The question becomes: can we come up with an algorithm that mimics a kind of moral discernment, one that says “this kind of sentiment is bad; avoid imitating it”? Will such an algorithm be enough? If moral discernment is where the line is drawn between an A.I., no matter how advanced, and a human being, then perhaps we should heed warnings like that of Stephen Hawking.
No comments:
Post a Comment