4min chapter

Towards Data Science cover image

104. Ken Stanley - AI without objectives

Towards Data Science

CHAPTER

The Creative Side of Intelligence?

i get enamoured with the sides of intelligence that are sort of really deep in what it means to be human. And i feel like that side of intelligence doesn't get as much, eh, or hasn't, atleast historically, gotten as much attention. creativity is basically about producing things, or generating things, that we didn't anticipate. So if you're getting unexpected things, then how do you say whether it's doing a good job? It brings up this issue of subjectivity. I think when you're evaluating creativityt questions have to be confronted. You can't just say, well, did you create something that's correct? Because who cares if nobody would merely want

00:00
Speaker 2
So maybe i'll open with a very general question. How do you think of intelligence? Like, what is intelligence to you? Yes,
Speaker 1
i think intelligence encompasses a whole range of different phenomena em and and i don't, actually, i'm not completely sure what all of them are, and it's an endless source of debate. But what i kind of get enamoured with are the sides of intelligence that are sort of really deep in what it means to be human, i guess, or the ones that move us emotionally. So that seems like things like, um, creativity or self expression, which are like, at the heart of the story of humanity, ike, you know, t the whole history of civilization is a historyof like, vast creativity. And there's these other things that we do, like we do calculations, we do analysis. We can, we can remember facts. We can control things em and those are very important, but that's not the side of it that em really amazes me and captivates me. It's really this creative, open handed side of it. And i feel like that side of intelligence doesn't get as much, eh, or hasn't, atleast historically, gotten as much attention. When we talk about i think of it more sortive, in the robotic sense, but i'm thinking of it in termsik the more deeply human sense. It also seems to be one of the hardest aspects
Speaker 2
of intelligence to evaluate as well, right? Like, i mean, i'm thinking here, when we look at classical use of e e ouno, vision classification and things like that, you can say, oh, well, you know your a, i performed to x % accuracy. Orr an entrope of y a. Whereas these kind of creative things are a bit different. Can you speak a little bit to to how that, how that plays into your thinking about intelligence? What's the difference between those things? Yes,
Speaker 1
yes. There's no doubtm, it's a good point that the creative side of intelligence, em it is hard to evaluate, in the sense that, how do you know, if you created some artificial a neural network or something like that, whether it's good at being creative? I mean, the problem with creativity is that it's basically about producing things, or generating things, that we didn't anticipate. I mean, if we did anticipate them, they sort of, like, not creative by definition. So if you're trying to make something that will make things that you don't anticipate, then you don't really have an objective ground truth of valuation for it. You don't know what you're valuating it against, cause you don't know what to expect. So if you're getting unexpected things, then how do you say whether it's doing a good job? And the problem is, it brings up this issue of subjectivity. I because there's a sort of, like, there's a subjective aspect of creativitys like, wha, do do i like it is a interesting m that's how i generally value it, creativey. Like, if you give me a story, im there's no objective way to say, it is or it isn't good, or, here's what the score is on a scale of one to a hundred. I just, how do i feel about that? Im even things like, ah, it's not that you may think, whell, there are, there's a side to creativity that's objective, like there's science and there's math, that you can write a proof, and it's either true or it's not true, or things like that. It's true, though, at the truth of the proof is actually a matter of objective fact. But what is still subjective is whether that's interesting, or is that worth proving, like, should you go down that route? And i think when you're evaluating creativityt questions have to be confronted. You can't just say, well, did you create something that's correct? Because who cares if nobody would merely want it or use it, or do anything with it or build on it? And so with this kind of research, you're grappling with, ultimately, something related to subjectivity, but trying to do it in as principled a way as to make it actually possibly algarithmic.

Get the Snipd
podcast app

Unlock the knowledge in podcasts with the podcast player of the future.
App store bannerPlay store banner

AI-powered
podcast player

Listen to all your favourite podcasts with AI-powered features

Discover
highlights

Listen to the best highlights from the podcasts you love and dive into the full episode

Save any
moment

Hear something you like? Tap your headphones to save it with AI-generated key takeaways

Share
& Export

Send highlights to Twitter, WhatsApp or export them to Notion, Readwise & more

AI-powered
podcast player

Listen to all your favourite podcasts with AI-powered features

Discover
highlights

Listen to the best highlights from the podcasts you love and dive into the full episode