Pieter Levels
speaker
696 appearances
1 recordings
1 series
first heard Aug 2024
last heard Aug 2024
Pieter Levels’s voice in public audio — every appearance, attributed to the second.
Trend
recordings per month · last 12 monthsNo recordings in the last 12 months.Older appearances are listed below; set an alert to hear about the next one.
Appearances
I had no problems, but I was very curious about technology and I was like playing with it and figuring out like, first just playing with it and then you find something like, okay, this generates, stable fusion generates houses very beautiful and interiors, you know?
Yeah, but that's very risky because that's the famous like solution trying to find a problem. Yeah. And usually it doesn't work. And that's very common with startup founders. I think they have tech, but actually people don't need the tech, right?
Yeah. So it's like fire your photographer. The idea is like you don't need a photographer anymore. You can train yourself as an AI model and you can take as many photos as you want anywhere in any clothes with facial expressions like happy or sad or poses, all this stuff.
So on the left, you have the prompt, the box. So you can write like, so model is your model. It's Lex Friedman. So you can write like model as a blah, blah, blah, whatever you want. Then press the button and it will take photos. It will take like one minute.
Replicate. Replicate.com.
Yeah. So I went to Google images, typed in Lex Friedman. I added like 10 or 20 images. You can open them in the gallery and you can use your cursor as to. Yeah. So some don't look like you. So the hit and miss rate is like, I don't know, let's say like 50-50 or something.
It was very bad in the beginning. It was so bad, but still people signed up to it, you know.
Man, but that's the problem with these models. No, we need to talk about this, because the models in Stable Diffusion, so the photorealistic models that are fine-tuned, they were all trained on porn in the beginning. And there was a guy called Hassan. So I was trying to figure out how to do photorealistic AI photos, and Stable Diffusion by itself is not doing that well. The faces look all mangled.
And it doesn't have enough resolution or something to do that well. But I started seeing these base models, these fine-tuned models, and people would train them on porn, and I would try them, and they would be very photorealistic. They would have bodies that actually made sense, like body anatomy.
But if you look at the photorealistic models that people use now still, there's still core of porn there, like of naked people. So I need to prompt out the naked, and everyone needs to do this with AI startups, with imaging. You need to prompt out the naked stuff. You need to put a, you know, naked... You have to keep reminding the model you need to put clothes on the thing.
Yeah, don't put naked because it's very risky. I have Google Vision that checks every photo before it's shown to the user to like check for NSFW.
Because you get the journalists get very angry if they, you know... There was a journalist, I think, that got angry that used this and was like, oh, it showed like a nipple because Google Vision didn't detect it. So there's like these kind of problems you need to deal with, you know? That's what I'm talking about. This is with cats. But look at the cat face. It's also kind of mangled.
You can zoom in on the cat if you want. Yeah.
It's a very sad cut.
But this, man, but this is the problem with AI startups because they all act like it's perfect. Like this is groundbreaking and, but it's not perfect. It's like really bad, you know, half the time.
Yeah. So you remove this stuff and you write like whatever you want, like in Thailand or something or in Tokyo.
Yeah. And then... You can say like at night with neon lights, like you can add more detail.
As a cowboy.
Man, I hope not. This is the end of my career.
Yeah, let's see.
Showing 121–140 of 696 · page 7 of 35
← Previous
Next →