This week, the limits to human lifespan.

We think that we're not going to be achieving any sort of radical increase in life expectancy anytime soon.

And when neural network artificial intelligence becomes stupid.

The state-of-the-art deep neural networks can be easily tricked into thinking that they're seeing

something when in fact what they're seeing is completely unrecognizable. Plus tracking mutations in

aging stem cells. This is the nature podcast for October the 6th, 2016. I'm Adam Levy. And I'm

Sharminy Bundell. This week, I've been finding out about problems with artificial intelligence, AI.

Ah, the inevitable rise of super intelligent robot overlords.

No.

Oh.

Well, not exactly.

Oh.

It's the problem of not knowing what a computer is thinking.

Our very own reporter, Davidei, Castelvechi, has written a feature on the topic this week, so I dragged him from his news desk to explain more.

So the article is about not so much what can we do with AI, but if AI gives us a certain answer, how do we know how it came to

that answer? So it's kind of can we trust the computers? Or there can be situations where we

know that they are giving the right answer, but we don't know why. And this is a particular,

particular kind of AI, isn't it?

Not just computers in general?

Correct.

So it's a kind of AI that has kind of taken the world by storm in the last four years or so.

It's called Deep Learning.

...