Gigaom AI Minute – February 23

Transcript

The range on when people think we will get an artificial general intelligence ranges from five to five hundred years. Generally speaking, people who think it is coming sooner are in the camp of people more likely to be concerned about it, and think that it's potentially a threat to humanity. Why do people disagree so much on this question? I think it boils down to five distinct reasons, and I'd like to cover each one of them in an AI Minute episode.

Number one, how far away are we from an unsupervised learner? The idea that you could take a machine learning system and essentially plug it into the internet and have it just ingest data and sort it and make sense of it without any human interaction has been a goal for a long time. It's the unsupervised model which says the machine isn't limited by the speed at which we can teach it. In essence, it can teach itself. People who think that artificial intelligence is coming sooner and is potentially more of a threat, believe we may be quite close to having an unsupervised learner. People who think it's far away and less of a threat also think that the unsupervised learner, generally speaking, is much further out if that is even possible.