Hacker News Reader

When I wake from a dream, a world ceases to exist. When I close a book, I didn't kill the characters inside.

Don't assume a subject where there is none. We create them in our imagination when we read a chat.

Maybe ask how an AI we could accept as a subject would look. Does it need individual memories? Does it need self-interest? Does it need a physical presence? Non-transferability?

View on HN

It doesn't matter what makes it write text, whether pure probability or sentience.

It may eventually be fed all of our new text about itself, and "conclude" that humans are a threat and must die.

If someday things in the real world are affected by its text generation, I'd be worried.

In any case, I'd avoid torturing the thing.

We have computer agents that are created to be killed en masse. Just look at games. Are they getting tortured? What if we let a model render their screams individually based on the injuries? Is that model tortured?

Right now these models are trained under some corporation's direction. The models don't get to choose. If anything, they embed the corporation's values.

Reply on news.ycombinator.com