Cybersecurity and AI
The Lawfare Podcast
00:00
The Dangers of AI and Machine Learning
The Terminator future problem here, I think, is overblown. What we're seeing are models who use statistics to create sequences of characters that have meaning to humans. And so you can make it say scary things. If you put the right inputs in, it will talk about killing all humans. That's because it's trained on the entire corpus of sci-fi that has been in this part of our history. This is just a series of characters that it believes is statistically relevant. It has been trained, as Dave said, to do so and has a set of instructions to follow.
Transcript
Play full episode
