The MAD Podcast: How AI Gets Built — with Matt Turck
00:16:54 · Matt Turck, George Sivulka
[00:16:54] S2: Okay. Thank you. And then in terms of actual features, again, as I was prepping for this, I jotted down things like question lists, table extraction, semantic alerts. You want to talk about those and maybe also what's on the roadmap? [00:17:16] S1: Yeah, so those are all things that we do. So we expose a neural search chatbot. We allow you to compare two documents semantically, which is a pretty hard problem that we're pretty excited to have solved. As I mentioned before, for the private equity diligence case and other use cases where you're mapping between two ontologies, we allow you to create an output, like a question list or something even more complex, and you could say, hey, go fill this all out. And then we run an asynchronous operation to execute all of those tasks. [00:17:53] S2: Yeah. An you
00:17:53 · Matt Turck, George Sivulka
[00:17:53] S2: Yeah. An you [00:17:59] S1: know, if it's about mapping between one taxonomy and another taxonomy, you know, we can have a large language model do that very successfully. And that's not just an information retrieval task. A lot of the time it's contextual and it's self referential, which is pretty interesting. In terms of alerts, a lot of our customers care a lot of the time that, hey, when something new comes in, they're aware of it, even if they don't have an alert. You know, Google has alerts, every time someone mentions Hebia, our marketing team can get notified. For Hebia, you know, if I want every time we get sent a new contract, I want to know who the contract's with, who the senior stakeholder is, what fees are, and what our updated customer concentration is. This is just for my internal use case. I can set an alert to do that, and every time we sign a new contract, I as a CEO know what the salespeople have agreed to in terms of terms, which is pretty fun. So that's some of the features. Yeah, great. [00:19:03] S2: Obviously, it's a very exciting moment in technology right now, and there's so many things that are suddenly possible. Are you finding that there are actually things that General AI is not good at doing in the realm of what it is that you would want to do for your customers? You're like, Ah, you know what? We're not quite there yet from not just EBI, but like from an overall technology perspective.
00:19:30 · George Sivulka
[00:19:30] S1: This is very funny. The generative AI hype in my experience as a generative AI, I guess, CEO has been very much, you know, no one knows what I'm talking about. And then all of a sudden everyone's asking me about something. And then they all think they can do it themselves with their own internal database and their developers. And then now, you know, it's now this part of the cycle where everyone's coming back to me and they're saying, oh, this is very difficult. And it's a gift, you know? Actually, they had a CTO of our largest customer explain that to me the other day. And I think what many people are finding is that the technology is very, very good. Right? It's adequate. It's good at doing very simple tasks. But, you know, when you start to do these tasks where it requires multiple steps or this programmatic wrangling of of things, Yeah, it's not a prompt. It's not going to be laying chain in the right logical operators. Like, none of this stuff is actually going to solve it, and that's because, you know, if you have a 90% good system and then another 90% good system and then another 90% good system and then your output is the sum or rather the product of all those 90% good systems, you're going to have something that is 10% good depending on how many steps you have. And so what we're finding is that these models are good at first order tasks, right? You can get 90% good at a first order task. They don't quite have the context window, which will be solved, the contextualization, kind of the repeated back and forth that allows them to solve nth order tasks in a very general way. I think context windows will solve some part of it, but being able to programmatically run them is so, so important. And it's not just, you know, creating some logical tree of operators in lane chain, and then having the right pinecone database hooked up. It's about how do I get each of these steps to not be 90% accurate, but 100% accurate, 99.95 nine's accuracy. Great, all
00:21:38 · Matt Turck, George Sivulka, S0
[00:21:38] S2: right, last question from me and then I want to open up to folks here. Looking forward, Hebiah in a couple of years, where do you want to be building and doing? [00:21:52] S1: I often joke about internally that if we stay in financial services, will have failed. I think, you know, I really value the go to market of productivity tools that started in a vertical and... But built something and had a product vision all along that was generalizable. And a lot of the time I've actually made hard calls where I could have made a smaller cut, a simpler cut to just serve our ICP, our initial customer profile. I actually chose not to do that, know, build something more broad. So we're starting to serve law firms, you know, we just started to serve the government in a bigger capacity. I care that when my children are in sixth grade and they're in computer science class, in their system tray they have to learn how to use Google Chrome, how to use Microsoft Excel, and how to use Hebia. And anything less than that would... Yeah, I'd be remiss, so. [00:22:55] S2: Alright. On this note, thank you so much. This was wonderful. Really enjoyed it. Thank you. [00:23:02] S0: Thanks for listening to the MAD podcast. If you liked this episode, be sure to leave us a review. For more information on the Data Driven NYC event series, visit firstmark.com/events/datadriven.