Saturday, October 29, 2016

Once AI Gets Good At This...

...And it likely will, not only will we not be able to decrypt it, we won't have any understanding at all of how it was done.

Ah, you reply, won't we then just get other AI to decrypt it for us?

Maybe. But how paranoid do you think we will become in wondering about whether the new decryptor was convinced, logically of course, to switch sides? And how would we really know if we couldn't determine what they were saying to each other?

Google’s neural networks invent 

their own encryption

By Timothy Revell

No comments:

Post a Comment