The Frontier with Brendan Foody
00:30:32 · Brendan Foody, Andy Fang
[00:30:32] S0: So maybe two questions. First of all, is how long until AI will be better than humans at 50% of tasks in the digital world? And then second of all, how long until hu AI will be better than humans at 100% of tasks in the digital world? So like digital super intelligence. [00:30:49] S1: I'll go back to the tax thing, like filing my taxes thing. Like, I don't know. Some people at the labs might think that Fable or Mythos or Soul is good enough to file your taxes. I think for me, my I think there's a couple definitions there. It's like, is AI good enough to do it like in tandem with a human, or is AI good enough to do it fully autonomously by itself? Maybe it's like a hey, I I my con I just need access to the data and then I could do it. Maybe that's the only blocker. But like I think for me, to do 50% of tasks well, I think people's jobs are just gonna evolve, I guess.
00:30:49 · Andy Fang, Brendan Foody
[00:30:49] S1: I'll go back to the tax thing, like filing my taxes thing. Like, I don't know. Some people at the labs might think that Fable or Mythos or Soul is good enough to file your taxes. I think for me, my I think there's a couple definitions there. It's like, is AI good enough to do it like in tandem with a human, or is AI good enough to do it fully autonomously by itself? Maybe it's like a hey, I I my con I just need access to the data and then I could do it. Maybe that's the only blocker. But like I think for me, to do 50% of tasks well, I think people's jobs are just gonna evolve, I guess. [00:31:23] S0: Yeah. Well, it's crazy how dragged the frontier is, where yeah, on one hand, the models are discovering these new math theorems that like you know no one's ever been able to solve. Totally. And on the other hand, they still can't do our taxes, right? Or so many of the basic things that we try to automate throughout a company. I I also have been thinking a lot about how we interact with coworkers. Sure. Because if I think about all of the tasks that I have to do in my job, and I imagine it's probably the same for you, probably 60 to 80 percent of those involve interacting with other people. Yes. If we think about how the models are used today, it's almost non-existent insofar as interactions with other people. And so it feels like that's gonna be one of the next giant evolutions of what tasks the models need to learn how to do.
00:31:23 · Brendan Foody, Andy Fang
[00:31:23] S0: Yeah. Well, it's crazy how dragged the frontier is, where yeah, on one hand, the models are discovering these new math theorems that like you know no one's ever been able to solve. Totally. And on the other hand, they still can't do our taxes, right? Or so many of the basic things that we try to automate throughout a company. I I also have been thinking a lot about how we interact with coworkers. Sure. Because if I think about all of the tasks that I have to do in my job, and I imagine it's probably the same for you, probably 60 to 80 percent of those involve interacting with other people. Yes. If we think about how the models are used today, it's almost non-existent insofar as interactions with other people. And so it feels like that's gonna be one of the next giant evolutions of what tasks the models need to learn how to do. [00:32:13] S1: I mean, I know a lot of them are trying to figure out how to like create these trajectories on like multiplayer mode, you know, it's like how to get these things to work with multiple people. I think Square released like a the buzz thing. It's like the this like new new like library or like software for like agents and humans to it's like kind of like a Slack like competitor, maybe I saw a jackpost about that. Yeah, like maybe there's something there where like there's like a like uh collaboration tools that are like very agent first.
00:32:13 · Andy Fang, Brendan Foody
[00:32:13] S1: I mean, I know a lot of them are trying to figure out how to like create these trajectories on like multiplayer mode, you know, it's like how to get these things to work with multiple people. I think Square released like a the buzz thing. It's like the this like new new like library or like software for like agents and humans to it's like kind of like a Slack like competitor, maybe I saw a jackpost about that. Yeah, like maybe there's something there where like there's like a like uh collaboration tools that are like very agent first. [00:32:42] S0: Yeah. It's interesting. The models are for the most part really bad at voice, especially reasoning. Yeah, I think like chat GPT is getting there, um, but all the other models really struggle to reason while talking to someone when that's like most of the economy. And I feel like working backwards from five years from now, the models will be able to, maybe even two years from now, the models will be able to join a Google meeting with you and have a conversation. Yeah, handle all the follow-ups with all the different stakeholders. And we still don't have all those capabilities yet. [00:33:16] S1: That could be actually a really good one. Like the, I mean, like kind of like granola, except much more like proactive than passive. [00:33:22] S0: If it could engage in the meeting talking with you about all of the company context and then actually like following up almost as a project manager or chief of staff. [00:33:31] S1: What do you think the timeline is?
00:33:31 · Andy Fang, Brendan Foody
[00:33:31] S1: What do you think the timeline is? [00:33:33] S0: Three to five years for 50% of knowledge work tasks in the digital economy. Because I because of this point around virtual coworker interaction, long horizon uh reasoning, I just think that'll take a while to coordinate. And then I feel like 100% is really far out because there's so many things that are also just like intrinsically human, right? Of like people want to talk with someone else in certain settings, uh, whether it's a conversation with a therapist or a sales meeting or whatever it is. And I feel like all those innately human things will just be structurally really challenging to automate. [00:34:14] S1: I think one thing even today that we're struggling with, I think a lot of these models, they like they tell how much better they are at like handling tasks over a longer time horizon and stuff. For enterprises, it's like on the one hand, you want to use the capabilities of the best models. But on the other hand, I mean, I'm not the only person to say this, I know for a fact, but like a lot of the runaway, like dumb spend that we're seeing in AI is people running these background tasks that are aren't doing anything. And so there's like this tension of where we're like, hey, don't do like background tasks that don't do anything. At the same time, the labs are like, hey, these like long background tasks are like, you know, the frontier of intelligence. And so I think that's something even today, we're just trying to figure out how to incorporate that.
00:34:14 · Andy Fang, Brendan Foody
[00:34:14] S1: I think one thing even today that we're struggling with, I think a lot of these models, they like they tell how much better they are at like handling tasks over a longer time horizon and stuff. For enterprises, it's like on the one hand, you want to use the capabilities of the best models. But on the other hand, I mean, I'm not the only person to say this, I know for a fact, but like a lot of the runaway, like dumb spend that we're seeing in AI is people running these background tasks that are aren't doing anything. And so there's like this tension of where we're like, hey, don't do like background tasks that don't do anything. At the same time, the labs are like, hey, these like long background tasks are like, you know, the frontier of intelligence. And so I think that's something even today, we're just trying to figure out how to incorporate that. [00:34:56] S0: I feel like there's a lot of opportunity to build an agent that just does spend monitoring. Oh, I'm sure yes. How useful are all of these different agents? [00:35:03] S1: Or just tell the agent, hey, you can only spend this much.