Episode 260: The Scream That Never Found a Voice (Murakami's "Sleep")

I'm a god.

You're a god?

I'm a god.

I'm not the god.

Don't think.

The greatest has spoken. Pay no attention to that man behind the curtain who are you? Who are you? A very bad man. I'm a very good man. Good man. They think deep thoughts and with no more brains than you have pay no attention to that man anybody can have a brain. You're a very bad man.

I'm a very good man just a very bad wizard welcome to Very Bad Wizards. I'm Tamela Summers from the University of Houston. Dave a woman had a full body orgasm during a performance of Tchaikovsky's Fifth Symphony by the La. Philharmonic. Dave, this is unbelievable. Did you know women could have orgasms?

That's an urban myth. That's what society tries to tell you.

Woke propaganda.

No woman I've ever been with, at least. It's pretty amazing. Like the coverage I just read a little bit of the coverage and people were like, nobody was shaming the woman. They were like, oh, and the orchestra kept playing and good for her, but I'm like, good for her. Unless her boyfriend was trying to finger her during that.

But I don't think that's what it was. You can tell if the people behind them and next to them would be right. I don't know.

Maybe she had one of those things that's like remote controlled.

She was wearing exactly, but occam's razor. It's the five Tchaikovsky's. Like, who among us won't splooce at the five, do you think?

Like to very bad wizards.

Episode Five Yeah, I love the ambulators. Did you listen to the four?

Yeah. People ask us all the time, do we have a central place where we list all the media that we've discussed? And the answer is no. So if anybody ever wants to build.

That out yeah, I don't know if this is someone did a letterbox of the movies and I don't know if that's still going, but I don't know either.

But it would be cool to have movies, short stories and books.

Among the episodes that would be listed, there would be this episode because in the second segment we will be diving for the first time into the work of Haruki Murakami. In particular, this episode, his short story Sleep, which is a banger. It is very good.

It is a banger. That's a good way to describe it.

People on Reddit always say that I call everything a masterpiece and I think everything is great. And it's well, like, we choose things because they're really good.

Totally. And we both spent quite a bit of time trying to find an appropriate story. So it's not like we're spinning the wheel and whatever it lands on, we call a masterpiece. But before we get to Murakami, we thought we'd talk a little bit about this came from a Tweet tweet. It was actually like months ago that I had put in the slack and not only is the Tweet from months ago, it's referring to a paper that's from a couple of years ago, from 2018 But I thought it was like a pretty clever idea given that all anybody ever wants to talk about nowadays is.

AI and how it's the only thing anybody wants to talk about.

And yet here we are. But this is a little twist on it. So this is a tweet from Ethan Malik, who was describing this paper. And the setup is basically this. You and an android are in front of a judge. The judge tells each of you to say one word. They will then kill whoever they think is the AI based on that. And so there's a paper from some researchers at MIT in the Journal of Experimental Social Psychology where they refer to this as the Minimum Turing Test. So one word is all you have to convince somebody that you're not an AI. What word would that be? Before we go, there's some results, but I think we should probably give our answers first.

Yeah. Yes. Why don't you go first? Because I realize I have thoughts about it, but I haven't fully on an actual word.

It's going to be controversial. I have gun to my head is the only way I would do it. Gun to my head, literally, I'm going to die. I would drop the N bomb.

I had that same thought and the.

Idea is simply that so many of these AIS that are out there have programmed explicitly into them not to use that word.

Right.

So I feel like if somebody says that word fully, then just probabilistically. It's not going to be an AI because it's hard to get the AI to say it.

Yeah. I wonder if that's a fair thing to do. Just assume that this AI has the constraints that the current chat bots or whatever have.

Yeah, it actually matters a lot like what you assume. When we get to a modified version of this, I'll talk a little bit more about why it matters to me not only what the AI is, but who you're telling it to. So in this case, we're told it's a judge, which I take it as just like some stand in for reasonable person.

And if it was someone I knew or whatever, that would make a huge difference. Totally. But I actually settled on so I was thinking about the N word. I wasn't thinking they had the constraint that the current ones have, but then I thought, well, these AIS typically get racist, but for whatever reason, I don't think that's true of cunt. So I think I would say that I don't see the AI saying that as just one word they could choose. And I feel like, yeah, I think.

The theme is whatever. AI just wouldn't to the extent that it's trying to appear human, it probably wouldn't go to the worst words and that's in fact, I feel like what these researchers found, which was that the most effective word, the word that most humans think would be most likely to tease you apart from an AI is poop.

Yeah. Did they have constraints of what they could say?

That's what I don't know, because I feel like just decorum would make it so that they wouldn't print anything worse than poop or they wouldn't give the words that we just gave as stimuli in the experiment.

So if you look actually, they have this chart where you see all the different ones of what the this is what AI thinks humans will pick. But compassion, soul, love seems like those two kinds of things, whereas fuck and poop and shit and penis and vagina and cunt. There's mine.

Oh, there it is.

Bootylicious. So funny. Is further away. So I think we're on the right track with this.

Right. What's interesting is that the way that these current AIS work with the language learning models, now that this is out there, they're just going to incorporate it. But I did earlier today, ask Chat GPT four, and it said the word I would use to convince someone that I am human, a human being and not an artificial intelligence is empathy. As a language model, I have a vast repository of knowledge and can provide helpful responses. But I lack the ability to truly feel and understand emotions like humans do. By using the word empathy, I am trying to convey that I recognize the importance of human emotions and connections, which is something that an AI may not fully understand.

This is what I don't get about what seems to be assumed here, is that we're trying to convince you with this word that you have the quality of the word. I don't think that's what if I was the judge, I wouldn't be looking and be like, oh, it must be human because it said empathy.

Yeah.

That's so stupid. Convince you I'm a bird by saying the word fly.

Yeah, right. I wouldn't think that. This is why I think you got to go more absurdist or contrary. You have to do something that doesn't make sense, perverse in the way that.

Paul talks about the perverse stuff. You have to put reason aside. I actually asked as a follow up question, I asked Chad GPT, what if you could use a whole sentence? What would it be? And it said, if I could use a whole sentence it always has to repeat your question like a very bad presenter. If I could use a whole sentence