[youtube]https://www.youtube.com/watch?v=bctMvKrB_y0[/youtube]
Computer programs can now spot liars better than humans听补苍诲 predict heart attacks four hours before a doctor. But artificial intelligence isn鈥檛 trying to compete with human minds: it reasons in a completely different way that can be beyond our comprehension.
Most successful artificial brains depend on machine learning, which relies on massive amounts of data to train an algorithm for a specific goal. The approach is allowing computers to successfully perform a variety of tasks, such as gauging a person鈥檚 mood or detecting cats in YouTube videos.
Advertisement
Probability is at the heart of this process. Initially, artificial intelligence doesn鈥檛 鈥渒now鈥� anything; it just assigns a probability to an outcome, such as the likelihood that a given video contains a cat. If it鈥檚 correct, it can then use that information to change the probability it assigns to the next video it encounters, and so on.
After enough cycles of guessing and receiving feedback, the algorithm will have a pretty good model of what a cat on a screen looks like.
Although we have a crude understanding of how an artificial mind works, the reasoning it uses as it crunches through a problem is often a mystery. Examining the code doesn鈥檛 provide insight into the internal logic that the machine uses.
A new technique akin to a brain scan for computers is aimed at finding out more by capturing snapshots of the process. Perhaps we will soon know as much about machine minds as they know about us.
is our new YouTube channel where we take the biggest ideas in science and make them easier to understand. so you don鈥檛 miss our latest episodes.
