Full episode with Michael Stevens (Dec 2019): https://www.youtube.com/watch?v=3qMemn__kK8
Clips channel (Lex Clips): https://www.youtube.com/lexclips
Main channel (Lex Fridman): https://www.youtube.com/lexfridman
(more links below)

Podcast full episodes playlist:
https://www.youtube.com/playlist?list=PLrAXtmErZgOdP_8GztsuKi9nrraNbKKp4

Podcasts clips playlist:
https://www.youtube.com/playlist?list=PLrAXtmErZgOeciFP3CBCIEElOJeitOr41

Podcast website:
https://lexfridman.com/ai

Podcast on Apple Podcasts (iTunes):
https://apple.co/2lwqZIr

Podcast on Spotify:
https://spoti.fi/2nEwCF8

Podcast RSS:
https://lexfridman.com/category/ai/feed/

Michael Stevens is the creator of Vsauce, one of the most popular educational YouTube channel in the world, with over 15 million subscribers and over 1.7 billion views. His videos often ask and answer questions that are both profound and entertaining, spanning topics from physics to psychology. As part of his channel he created 3 seasons of Mind Field, a series that explored human behavior.

Subscribe to this YouTube channel or connect on:
- Twitter: https://twitter.com/lexfridman
- LinkedIn: https://www.linkedin.com/in/lexfridman
- Facebook: https://www.facebook.com/lexfridman
- Instagram: https://www.instagram.com/lexfridman
- Medium: https://medium.com/@lexfridman
- Support on Patreon: https://www.patreon.com/lexfridman

Let's jump back to it, artificial intelligence.

What's your thought of the state of where we are at currently with artificial intelligence and what do you think it takes to build human level or superhuman level intelligence?

That's my biggest question at the moment.

And I think it's because my instinct is always to go, well, what are the foundations here of our discussion?

How do we measure the intelligence of an artificial machine or a program or something?

Because there's also a fascinating field of how do you measure human intelligence.

But if we just take that for granted, saying that whatever this fuzzy intelligence thing we're talking about, humans kind of have it.

So Turing developed a test that's natural language conversation.

A chat bot that you'd want to hang out and have a beer with, you know, for a bunch of hours or have dinner plans with, is that a good test, natural language conversation?

Is there something else that would impress you?

If there was a chatbot that was like incredibly, I don't know, really had a personality.

Like if I'm unable to tell that it's not another person, but then I was shown a bunch of wires and mechanical components and it was like that's actually what's you're talking to I don't know if I would feel that guilty destroying it

would feel guilty because clearly it's well made and it's a really cool thing It's like destroying a really cool car or something.

But I would not feel like I was a murderer.

So yeah, at what point would I start to feel that way?

And this is such a subjective psychological question.

If you give it movement, or if you have it act as though, or perhaps really feel pain as I destroy it and scream and resist, then I'd feel bad.

So if you just have a robot that not screams, just like moans in pain, if you kick it, that immediately just puts it in a class that we humans, it becomes, it anthropomorphize it, it almost immediately becomes human.

So that's a psychology question as opposed to sort of a physics question.

Right, I think that's a really good instinct to have.

Screams and moans, even if you don't believe that it has the mental experience, the qualia of pain and suffering, I think it's still a good instinct to say, you know what, I'd rather not hurt it.

The problem is that instinct can get us in trouble because then robots can manipulate that.

There's robots like the Facebook and the YouTube algorithm that recommends the video, and they can manipulate it in the same kind of way.


Speaker 0: Let's jump back to it, artificial intelligence. What's your thought of the state of where we are at currently with artificial intelligence and what do you think it takes to build human level or superhuman level intelligence?

Speaker 1: I don't know what intelligence means. That's my biggest question at the moment. And I think it's because my instinct is always to go, well, what are the foundations here of our discussion? What does it mean to be intelligent? How do we measure the intelligence of an artificial machine or a program or something?

Speaker 0: Can we say that humans are intelligent? Because there's also a fascinating field of how do you measure human intelligence. Of course. But if we just take that for granted, saying that whatever this fuzzy intelligence thing we're talking about, humans kind of have it. What would be a good test for you?

So Turing developed a test that's natural language conversation. Would that impress you? A chat bot that you'd want to hang out and have a beer with, you know, for a bunch of hours or have dinner plans with, is that a good test, natural language conversation? Is there something else that would impress you? Or is that also too difficult

Speaker 1: to think about?

Speaker 0: Oh yeah,

Speaker 1: I'm pretty much impressed by everything. I think that if- Roomba? If there was a chatbot that was like incredibly, I don't know, really had a personality. And if I didn't, the Turing test, right? Like if I'm unable to tell that it's not another person, but then I was shown a bunch of wires and mechanical components and it was like that's actually what's you're talking to I don't know if I would feel that guilty destroying it

Speaker 0: I

Speaker 1: would feel guilty because clearly it's well made and it's a really cool thing It's like destroying a really cool car or something. But I would not feel like I was a murderer. So yeah, at what point would I start to feel that way? And this is such a subjective psychological question. If you give it movement, or if you have it act as though, or perhaps really feel pain as I destroy it and scream and resist, then I'd feel bad.

Speaker 0: Yeah, it's beautifully put. And let's just say act like it's a pain. So if you just have a robot that not screams, just like moans in pain, if you kick it, that immediately just puts it in a class that we humans, it becomes, it anthropomorphize it, it almost immediately becomes human. So that's a psychology question as opposed to sort of a physics question.

Speaker 1: Right, I think that's a really good instinct to have. You know, if the robot- Screams. Screams and moans, even if you don't believe that it has the mental experience, the qualia of pain and suffering, I think it's still a good instinct to say, you know what, I'd rather not hurt it.

Speaker 0: The problem is that instinct can get us in trouble because then robots can manipulate that. And there's different kinds of robots. There's robots like the Facebook and the YouTube algorithm that recommends the video, and they can manipulate it in the same kind of way.

See all Lex Fridman transcripts on Youtube

Vsauce: What Does it Mean to Be Intelligent? | AI Podcast Clip with Michael Stevens