The Information's TITV
00:42:57 · Akash Pasricha, Andrew Ambrosino
[00:42:57] S0: And and as you think about costs, I mean, is kinda something I've been thinking about. We kind of equate it with models and tokens and compute all the time and efficiency. I mean, I wonder if there are product level decisions in the, you know, the user experience that you can make to make it cheaper, or make things more efficient? Not necessarily the model or the compute, but literally, like, the product itself. [00:43:20] S5: Yeah. Certainly. And we do we do some of this with, you know, the modes and, like, the reasoning selector and things like this. I think, whether everyone should be choosing reasoning and models every time... You know, a lot of people love to do this who get a, you know, sense for development, but some people want... They they want the right thing being chosen. Right? The most... Yeah.
00:43:20 · Andrew Ambrosino, Akash Pasricha
[00:43:20] S5: Yeah. Certainly. And we do we do some of this with, you know, the modes and, like, the reasoning selector and things like this. I think, whether everyone should be choosing reasoning and models every time... You know, a lot of people love to do this who get a, you know, sense for development, but some people want... They they want the right thing being chosen. Right? The most... Yeah. [00:43:45] S0: Right. I'll end with this. I mean, you know, I've been thinking a lot about how we are going to teach people the capabilities of these tools because this is the conversation that everyone is having. The models can do so much, and yet people aren't even awake Right. To that right now. But then I come back to my own experience with it, is that I'm... Like, I was never a software developer. I'm not that technical a person. I still get a little freaked out when I'm trying to use the tools, I see it doing stuff that I don't understand. Not from a security perspective, just like I literally don't understand what is being made in front of my eyes. So how do you think that we are going to teach people to use these tools in a more advanced way when you have someone like me who is a pretty tech forward person, but still, if I don't understand what it's doing in front of my eyes, I I walk away from it a little bit.
00:44:42 · Andrew Ambrosino
[00:44:42] S5: Alright. So I have a take on this. I I have a a new take. New take? New to me. A new new take to me. And if you remember in 2007 when the iPhone came out, all of the apps were, like, leather and paper and glass. Like, they looked like the materials. Right? They looked like a notebook. And this... It's called skeutomorphism. Right? And part of the the the reason why that was the right move at the time and that we don't have that anymore is because it helped guide people from one world to the other. Right? So it said, know, this is a familiar calculator. It's a familiar notebook. And I think that there are two things that are very similar in what we work on that I kind of view in the same lens. One is cursors. Not the company, but the the cursors on your screen. Right? Computer use, as is in the ChatGPT app, does not need to show you a cursor. Doesn't It need it for itself. Browser use doesn't need a cursor. When when the agent modifies a space page, it doesn't need a cursor. Right? This is all so that it feels familiar and that it's easy to track the work that's happening. Right? So this is one area of of sort of skeetomorphism. And then the other one is in named agents like DOTS. Right? As many people have pointed out, like, you can do a lot of what DOTS are are giving people. You could ask Codex to do it. You could ask ChatGPT work to do it. But in embodying and personifying these agents, it... For a lot of people, especially at work when they're used to working with teams of people, has really helped on the what what can I do with this thing? And it's, you know, making it feel like a coworker is is helping with this. Right? The, like, oh, yeah. Can be proactive. I can ask it to remind me of things. It kind of helps bridge that gap in the short term even though it may not be strictly necessary to to have that form. So I found this quite interesting, actually.
00:46:28 · Akash Pasricha, Andrew Ambrosino, Dan Shipper
[00:46:28] S0: But let me just ask you the flip side to that, which is that you talk about personifying these agents. Yes. I mean, there is kind of a whole school of thought that we shouldn't be personifying them too much and treating them like humans because of the fact that, you know, as these capabilities get better, that could be a risk. So, you know, where where do you sort of sit on that balance? [00:46:49] S5: Yeah. I I understand this take. I think with dots, you're seeing us embody but not personify, which is, you know, kind of a little bit nuanced. And, you know, I think that we've done a very good job with that product, we'll iterate, and we'll see how, you know, these things, these things will surely evolve over time. [00:47:08] S0: Right. [00:47:09] S5: But I think that there's there's a a, you know, threading that has to happen here. [00:47:13] S0: Yeah. Great. Well, Andrew, I wanna thank you for Thank you. Joining us. That is Andrew Ambrosino from OpenAI here on TITV. I'm here with Dan Shipper, the CEO and founder of Every. Dan, welcome to the show. It's great to have [00:47:40] S5: you here. [00:47:41] S2: Thanks for having me, and thanks for coming to Thesis. [00:47:43] S0: Wait. Is this like your Super Bowl? [00:47:44] S2: Like Yes. In fact... Well, you know, GPT 8 probably would be, like, my my true Super Bowl, but barring that, this this is a really special gathering, and it's it's really great to have
00:47:44 · Dan Shipper, Akash Pasricha
[00:47:44] S2: Like Yes. In fact... Well, you know, GPT 8 probably would be, like, my my true Super Bowl, but barring that, this this is a really special gathering, and it's it's really great to have [00:47:57] S0: you here. Well, it's a great event. I'm excited to talk about all that's news right now in AI. Tell me, what's the most counterintuitive thing you've heard on the ground here today at the event about AI? See, I don't know if [00:48:08] S2: this is counterintuitive, but I think it's I think it's really interesting, and it says something about where things are going. Because everybody at this event... If you're here, there's 400 people here. If you're here, you love AI. And I've talked to so many people at this event who are working in big companies who are getting so much pressure to use AI as much as possible, which is... They love. It's good for them. It's it's like... It's what it's what they want. And I think the expectations from management or executives that are not familiar with AI for how fast you can go before you set up a good system to do it well, there... It's like... It's kind of out of whack. And so there's something really interesting where for AI pill people inside of big companies, the way that they're being managed, I think, is is out of whack with, like, what is realistic and and the and the the best way to start a high quality AI project, which is to get the basics right first and build, like, build slowly, and then you can go fast. Go slow to go
00:48:08 · Dan Shipper, Akash Pasricha
[00:48:08] S2: this is counterintuitive, but I think it's I think it's really interesting, and it says something about where things are going. Because everybody at this event... If you're here, there's 400 people here. If you're here, you love AI. And I've talked to so many people at this event who are working in big companies who are getting so much pressure to use AI as much as possible, which is... They love. It's good for them. It's it's like... It's what it's what they want. And I think the expectations from management or executives that are not familiar with AI for how fast you can go before you set up a good system to do it well, there... It's like... It's kind of out of whack. And so there's something really interesting where for AI pill people inside of big companies, the way that they're being managed, I think, is is out of whack with, like, what is realistic and and the and the the best way to start a high quality AI project, which is to get the basics right first and build, like, build slowly, and then you can go fast. Go slow to go [00:49:11] S0: fast. And this is kind of interesting though because you're saying management's expectations are too high, and yet we've got headlines every week saying x y z company is putting a cap on how much you you you spent on AI.