Adam Leventhal

I like when chat GPT is very wrong, I just keep at it. I just, I'm like, no chat GPT. How about a little bit of this? How about a little of that? And I get in arguments with it. And it's like, this is a two dimensional image. I'm like, chat GPT, you have given me a three dimensional image. Like, I don't, I don't know how to like, we're kind of at loggerheads.

And then I've had chat GPT say like, basically let's agree to disagree. Like just kind of using its managerial techniques to, to get me to back off. Yeah.

And then I've had chat GPT say like, basically let's agree to disagree. Like just kind of using its managerial techniques to, to get me to back off. Yeah.

So then I downloaded not one, but several physics simulators. I finally got one to work. I think it's built in Java. I drew the thing. I got some image that looked okay, downloaded it to my son Will's iPad where he has an Apple Pencil, and I traced it with the Apple Pencil to get the XKCD style back onto the computer. This is all during an all-hands and stuff. I'm trying not to have this...

So then I downloaded not one, but several physics simulators. I finally got one to work. I think it's built in Java. I drew the thing. I got some image that looked okay, downloaded it to my son Will's iPad where he has an Apple Pencil, and I traced it with the Apple Pencil to get the XKCD style back onto the computer. This is all during an all-hands and stuff. I'm trying not to have this...

He listens to this podcast. He may be a little... I mean, yeah. During an all-hands that I was paying... Hey, look, I wasn't doing a crossword puzzle or something.

He listens to this podcast. He may be a little... I mean, yeah. During an all-hands that I was paying... Hey, look, I wasn't doing a crossword puzzle or something.

Exactly. And then it just came out great. I was really pleased with how it worked out. So great, in fact, that I've seen it like... used in other people's podcasts, which was a little surprising. And then I got a little uppity about like, hey, how dare you steal our images? And then I looked at all of our other images, which is like mostly me stealing copyrighted material.

Exactly. And then it just came out great. I was really pleased with how it worked out. So great, in fact, that I've seen it like... used in other people's podcasts, which was a little surprising. And then I got a little uppity about like, hey, how dare you steal our images? And then I looked at all of our other images, which is like mostly me stealing copyrighted material.

So it's a little late in the day for me to be complaining.

So it's a little late in the day for me to be complaining.

for the image but i feel like i mean that maybe it's a it's an outlier but it's not an outlier by that much i feel we've it's an outlier yeah yeah but it's also like if if it took an hour and a half that would be a lot like i think it's probably closer to an hour so it really is like

for the image but i feel like i mean that maybe it's a it's an outlier but it's not an outlier by that much i feel we've it's an outlier yeah yeah but it's also like if if it took an hour and a half that would be a lot like i think it's probably closer to an hour so it really is like

I do have one favorite of my images, which are we going to talk about the open AI boardroom brawl? Oh, was that what that was? That was last year.

I do have one favorite of my images, which are we going to talk about the open AI boardroom brawl? Oh, was that what that was? That was last year.

What I like from this year was open source LLMs with Simon Willison, which also was... That was a great get on your part again. But the image is like a robot reading the New York Times. And a big theme of that episode was the provenance of the training data. And like... look, this is all this copyrighted material.

What I like from this year was open source LLMs with Simon Willison, which also was... That was a great get on your part again. But the image is like a robot reading the New York Times. And a big theme of that episode was the provenance of the training data. And like... look, this is all this copyrighted material.

So I do love that image of the AI casually reading the New York Times as it obviously memorizes it and then feeds it back to you.

So I do love that image of the AI casually reading the New York Times as it obviously memorizes it and then feeds it back to you.

Yeah, as someone told us, they referred to, they told us their child referred to this as grand theft autocomplete, which I thought was awesome. That is, that's, man, that's, is that a teenager? Yeah, we were there. We were at dtrace.conf when Alex was telling us that, I think.