Menu
Sign In Search Podcasts Charts People & Topics Add Podcast API Pricing

Dax

👤 Person
3156 total appearances

Appearances Over Time

Podcast Appearances

How About Tomorrow?
How Dangerous Can AI Get, Dax is Down on DeepSeek, and AI First App Development

They talked about their techniques, and their techniques are interesting and novel, so you can't take that away from them. But then they separately claim that we use these techniques on this hardware. to achieve this outcome. But there's so many ways to lie about that.

How About Tomorrow?
How Dangerous Can AI Get, Dax is Down on DeepSeek, and AI First App Development

They talked about their techniques, and their techniques are interesting and novel, so you can't take that away from them. But then they separately claim that we use these techniques on this hardware. to achieve this outcome. But there's so many ways to lie about that.

How About Tomorrow?
How Dangerous Can AI Get, Dax is Down on DeepSeek, and AI First App Development

The thing is, if you... Okay, let's say someone told you that, hey, I can... run a SQL query that filters a trillion rows in half a second. right? You, as someone that understands this stuff, you're like, I'm not even going to waste my time reproducing that because that makes no sense. Okay. Yeah.

How About Tomorrow?
How Dangerous Can AI Get, Dax is Down on DeepSeek, and AI First App Development

The thing is, if you... Okay, let's say someone told you that, hey, I can... run a SQL query that filters a trillion rows in half a second. right? You, as someone that understands this stuff, you're like, I'm not even going to waste my time reproducing that because that makes no sense. Okay. Yeah.

How About Tomorrow?
How Dangerous Can AI Get, Dax is Down on DeepSeek, and AI First App Development

Yeah, I mean, just because it's such a hyped space, it's so hard to tell what's real and what's not. And also, the noise comes on both sides. So remember that we're saying novel because it's been published publicly. We don't know that OpenAI already ran across this and is using it to develop their stuff, the techniques in there.

How About Tomorrow?
How Dangerous Can AI Get, Dax is Down on DeepSeek, and AI First App Development

Yeah, I mean, just because it's such a hyped space, it's so hard to tell what's real and what's not. And also, the noise comes on both sides. So remember that we're saying novel because it's been published publicly. We don't know that OpenAI already ran across this and is using it to develop their stuff, the techniques in there.

How About Tomorrow?
How Dangerous Can AI Get, Dax is Down on DeepSeek, and AI First App Development

So this might not even be a surprise to them as much as be, oh, they like independently came across the same techniques and they know that, yeah, it's not causing like a thousand X decrease in draining costs. But then the other noises, and so now, and this is a part where I'm like, okay, this could be noise from the other side, but I did think about this when it came out.

How About Tomorrow?
How Dangerous Can AI Get, Dax is Down on DeepSeek, and AI First App Development

So this might not even be a surprise to them as much as be, oh, they like independently came across the same techniques and they know that, yeah, it's not causing like a thousand X decrease in draining costs. But then the other noises, and so now, and this is a part where I'm like, okay, this could be noise from the other side, but I did think about this when it came out.

How About Tomorrow?
How Dangerous Can AI Get, Dax is Down on DeepSeek, and AI First App Development

OpenAI is claiming that they have proof that DeepSeq was trained on outputs of their models or of like some maybe potentially like unauthorized access to stuff from OpenAI. Okay. And there's like some like, again, this doesn't mean anything, but like.

How About Tomorrow?
How Dangerous Can AI Get, Dax is Down on DeepSeek, and AI First App Development

OpenAI is claiming that they have proof that DeepSeq was trained on outputs of their models or of like some maybe potentially like unauthorized access to stuff from OpenAI. Okay. And there's like some like, again, this doesn't mean anything, but like.

How About Tomorrow?
How Dangerous Can AI Get, Dax is Down on DeepSeek, and AI First App Development

like the pseudoscience part of this is people were able to get DeepSeek to reply and make the exact same mistakes that O1 makes, which seems like maybe it's a coincidence. Maybe it means something. But yeah, the point here is like, it's just such a crazy hype space with a ton of money that there's like zero ability to draw any kind of, this is what's happening right now in the moment.

How About Tomorrow?
How Dangerous Can AI Get, Dax is Down on DeepSeek, and AI First App Development

like the pseudoscience part of this is people were able to get DeepSeek to reply and make the exact same mistakes that O1 makes, which seems like maybe it's a coincidence. Maybe it means something. But yeah, the point here is like, it's just such a crazy hype space with a ton of money that there's like zero ability to draw any kind of, this is what's happening right now in the moment.

How About Tomorrow?
How Dangerous Can AI Get, Dax is Down on DeepSeek, and AI First App Development

It's just impossible for situations like this.

How About Tomorrow?
How Dangerous Can AI Get, Dax is Down on DeepSeek, and AI First App Development

It's just impossible for situations like this.

How About Tomorrow?
How Dangerous Can AI Get, Dax is Down on DeepSeek, and AI First App Development

You know how Sam Altman is. He just did the whole, like, generic, wow, it's really impressive, and I'm invigorated by the competition. You know, just like the fucking, he, to be honest, ChatDBT is more human than Sam Altman already. Yeah. Did you see what Claude did to me yesterday? No, what? Did you tweet about it or something? I can't believe it did this.

How About Tomorrow?
How Dangerous Can AI Get, Dax is Down on DeepSeek, and AI First App Development

You know how Sam Altman is. He just did the whole, like, generic, wow, it's really impressive, and I'm invigorated by the competition. You know, just like the fucking, he, to be honest, ChatDBT is more human than Sam Altman already. Yeah. Did you see what Claude did to me yesterday? No, what? Did you tweet about it or something? I can't believe it did this.

How About Tomorrow?
How Dangerous Can AI Get, Dax is Down on DeepSeek, and AI First App Development

So I was trying to deal with, again, bringing it all back down to earth. I was trying to insert something into a Postgres database. And of course, on conflict, you want to do an update operation. Of course. I'm used to my SQL where you can just say on any conflict, do this operation. But on PostgreSQL, it specifies like, oh, when this conflicts, do that. When that conflicts, do this.

How About Tomorrow?
How Dangerous Can AI Get, Dax is Down on DeepSeek, and AI First App Development

So I was trying to deal with, again, bringing it all back down to earth. I was trying to insert something into a Postgres database. And of course, on conflict, you want to do an update operation. Of course. I'm used to my SQL where you can just say on any conflict, do this operation. But on PostgreSQL, it specifies like, oh, when this conflicts, do that. When that conflicts, do this.

How About Tomorrow?
How Dangerous Can AI Get, Dax is Down on DeepSeek, and AI First App Development

But I was like, OK, can I just on conflict on anything? Is that possible? And Claude, in a single reply, writes out, hey, yeah, you can do this. And then it writes out the query. And then right after it does that, it continues writing, being like, just kidding. That syntax doesn't exist. What? It said just kidding? Oh, my God. This tweet has 5,000 likes. I didn't even notice. What?

How About Tomorrow?
How Dangerous Can AI Get, Dax is Down on DeepSeek, and AI First App Development

But I was like, OK, can I just on conflict on anything? Is that possible? And Claude, in a single reply, writes out, hey, yeah, you can do this. And then it writes out the query. And then right after it does that, it continues writing, being like, just kidding. That syntax doesn't exist. What? It said just kidding? Oh, my God. This tweet has 5,000 likes. I didn't even notice. What?