May 13, 2026 · 0 min · 27 segments
We discuss the rise of Generative AI in game development, with Darren Grey, Jeremiah Reid and Tommy Thompson. Read more »
No entities detected.
So the way that I usually distill it is if you think historically, we've talked a lot about machine learning.
So there's kinda, I, I would say there's three disciplines of AI, right? Well, generative AI is technically one of the, one of the two.
But, um, so I- we typically think of old school AI as symbolic in nature, where we define a bunch of rules and logic around how a pro- a problem works, and then from that we have agents.
These are systems that would exist within the thought space and the logic that we have defined to search for answers.
And, you know, even traditional procedural content generation, I would consider kind of falls along that, those lines, right? We define the logic of our, our content generation systems do, we let them execute within that space, often with a bit of randomness, and boom, we get an output.
So a lot of traditional, like, um, AI, NPCs and stuff like that in games, we still use a lot of symbolic systems and kind of these kind of search and logic-based approaches.
Then you have a lot of the kind of contemporary conversation around AIs around machine learning, where you're training a model against the inputs and outputs.
You're usually taking data from a problem and then getting it to try and replicate that problem such that it can understand it to a certain degree.
So, you know, you think about image recognition, you take a bunch of pictures, you get it to recognize what those images are, cats, dogs, et cetera, is the example I love to use.
I give it a picture of a cat and it says, "I'm 97% confident that's a cat." And it's like, okay, cool.
So traditional machine learning is always about predicting answers to a problem, whereas generative AI is a kinda specific subset of machine learning where you now get it to generate outputs evocative of the input data.
So before, I was trying to get it to guess whether or not it's a cat or a dog.
Now with a generative model, I would get it to generate an image of a cat or a dog based on the images that I've given it.
And when you think about it in, in the kind of, in the general, you know, kind of public consciousness, that's when it comes to things like GPT and stable diffusion, Sora, all these other kind of generative systems that are creating content based on the information that it's given.
So we're always, it's always about training a model that's designed to kind of reproduce features of the things that it has seen, that it has learned from.
That's, that's usually how I try and distill it, whi- while keeping, like, some of the, I don't know, the scientific rigor of it there to some degree.
So the way that I usually distill it is if you think historically, we've talked a lot about machine learning.
So there's kinda, I, I would say there's three disciplines of AI, right? Well, generative AI is technically one of the, one of the two.
But, um, so I- we typically think of old school AI as symbolic in nature, where we define a bunch of rules and logic around how a pro- a problem works, and then from that we have agents.
These are systems that would exist within the thought space and the logic that we have defined to search for answers.
And, you know, even traditional procedural content generation, I would consider kind of falls along that, those lines, right? We define the logic of our, our content generation systems do, we let them execute within that space, often with a bit of randomness, and boom, we get an output.
So a lot of traditional, like, um, AI, NPCs and stuff like that in games, we still use a lot of symbolic systems and kind of these kind of search and logic-based approaches.
Then you have a lot of the kind of contemporary conversation around AIs around machine learning, where you're training a model against the inputs and outputs.
You're usually taking data from a problem and then getting it to try and replicate that problem such that it can understand it to a certain degree.
So, you know, you think about image recognition, you take a bunch of pictures, you get it to recognize what those images are, cats, dogs, et cetera, is the example I love to use.
I give it a picture of a cat and it says, "I'm 97% confident that's a cat." And it's like, okay, cool.
So traditional machine learning is always about predicting answers to a problem, whereas generative AI is a kinda specific subset of machine learning where you now get it to generate outputs evocative of the input data.
So before, I was trying to get it to guess whether or not it's a cat or a dog.
Now with a generative model, I would get it to generate an image of a cat or a dog based on the images that I've given it.
And when you think about it in, in the kind of, in the general, you know, kind of public consciousness, that's when it comes to things like GPT and stable diffusion, Sora, all these other kind of generative systems that are creating content based on the information that it's given.
So we're always, it's always about training a model that's designed to kind of reproduce features of the things that it has seen, that it has learned from.
That's, that's usually how I try and distill it, whi- while keeping, like, some of the, I don't know, the scientific rigor of it there to some degree.
The rest of this transcript — segmented and speaker-labeled, so you land on the exact moment something was said
Search every transcript — by keyword, by phrase, or by meaning, across every show Radar indexes
Trends — what is surging across podcasts, measured against its own baseline
Alerts — when a name you follow appears in a newly indexed episode
No account is needed to search Radar.