Pinker: We don't know the limits of AI

"We don't yet know its limits."

Harvard Psychology Professor Steven Pinker talks AI.

[posthocprior]

This interview should have been much longer. I make predictive models, for my job, and this is one of the best descriptions I’ve heard of his machine learning algorithms work.

[auditoryproductions]

4:14 He makes a good point that it's an alien intelligence in a sense

[vavaleo]

A word of advice:
Build an A.I. machine with an on/off switch.
You never know when things are going to go wrong.

[DavidSwanson]

Steven: It doesn't have common sense. It doesn't know, for example, that if someone was alive at 9am and alive at noon, that must mean they were alive at 10am.
Query to ChatGPT: If someone was alive at 9am and 12pm, what can we say about them at 10am?
ChatGPT: If someone is alive at 9 am and 12 pm, we can infer that they were alive at 10 am as well.

[oystercatcher]

Interesting. But my knowledge is spread across the billions of neurons in my brain. No one can open my brain and see what I know. But they can test me with questions and use interviews and exams. This is just the same for AI. So the limitations are not so different

[jonassteger-jensen]

i asked Bard: "if a man was alive at 9 am and alive at noon - was the man alive at 10 am?"

Bard answered: "Yes, it is logical to assume that the man was alive at 10 am if he was alive at 9 am and noon. There is no information to suggest that he died at any point between 9 am and 10 am, or between 10 am and noon. Therefore, the most likely explanation is that he remained alive throughout this time period."

So that was just a extremely bad example. Here it actually masters common sense.

[RobertSela]

This guy didnt blink once, its amazing. Has he been replaced by an AI robot?

[ssuwandi]

Ohh we know. AI is incapable to respond "I DON'T KNOW"

[reallifefaith]

It's so funny to listen to guys like this insist that there is something special about human consciousness while denying that the world is anything but physical.