Proof of Work: AI Value Creation
00:36:13 · Joe Zein
[00:36:13] S2: Depends on on the plan you're on, right? It's also, you know, uh the enterprise plans cover for that, right? You you ChatGPT Enterprise and Claude, I think it's Team Plan Plus, well, you know, in the agreement is not using your data for training, for future training, and and obviously you're operating in an insulated environment. That being said, you're trusting those companies, sure, I think. And by the time they retrain and that data, if there's a leak, I mean, again, it's a it's it's a trust game, right? And it's very mission critical. Like like when you see someone or when you see the government going and using those models, you I think I I personally better believe that that that you know there's no cross-contamination across orics. But I'm a I know I'm I'm an optimist about that, about that stuff. I've seen people wanting to go full open source, um, because on on open source models, you're hosting the model on a machine that you own, so you know no one is going to train it. That data's not going anywhere and it's seeing within your network, within your infrastructure. Uh, so that's the answer for people who are extremely cautious about their data and don't trust the frontier model provider companies. Uh they decide to go with something you know open source that they host, but then there's also cost associated with that. There's uh maintenance associated with that. Uh, so it's you know, pros and cons.
00:37:31 · Yann Magnan
[00:37:31] S1: Yeah, I would uh answer something uh quite similar, and I think it depends a lot on uh what type of data you we're talking about, what strategy uh you uh are addressing as a private capital fund, maybe a credit fund doesn't have the same uh uh sensitivity on confidentiality data as a private equity fund that uh invests in uh defense uh assets, just to give an example. Uh so depending on what we do you do, you'll have a different sensitivity to that question. I I'm I'm an optimist as well, same as Joe, so I would uh I tend to believe that as long as you activate these enterprise uh licenses, uh you should be safe. That being said, in our platform, we can turn off and off the LLMs depending on whether the clients are agreeing to that or not. The last comment I would make here is that all of us have been uh using uh LLMs. We've seen um you know the uh improvements uh that they that that that have that that they have made in the past few quarters. I guess we can all guess that the what's coming ahead is going to be like pretty, pretty amazing. And the question as well that I that I asked to my clients before they make this decision, LLMs or not, is are you willing to potentially pass on that intelligence that these LLMs are going to provide you with because your data is too confidential? If you're and if the answer is yes, that's fine, fully understandable. If you think that your data is much is so important that you you'd prefer pass passing on the uh extra layer of intelligence and and keeping your data in a place where you're certain that it's safe, that's fine. That's a good answer. But you have to be you have to be conscious that you'll be passing on massive intelligence, and your competitors might not, by the way. And and you have to think about that.
00:39:18 · Stuart Willson, Yann Magnan
[00:39:18] S0: Yeah. Uh as we come to a close, I have a few questions, but I want to ask you two. Is there something critical that the audience should know about the intersection of AI and portfolio monitoring that we haven't yet discussed? Anything you want to share with the audience before I get to my closing questions? [00:39:39] S1: Uh I'll give it a start. I I would say what we've discussed today. We have the same discussion in six months' time, and we'll we will be in a different world. Things are moving like really, really fast. Really, really fast.
00:39:51 · Joe Zein
[00:39:51] S2: Yeah. Yeah, things are, I think I I agree. I think I think things are like even for us, you know, where we're we we work with this stuff day in, day out, and and we, you know, we've barely been catching up. I think people still need to be a little bit cautious because now building a UI is extremely easy, building a system that works reliably uh is not, and a lot of people have been trying to build skyscrapers on top of wooden foundations. So that's also probably the time to really focus on solidifying your foundations because I don't see as fast as we're going, I don't see a world where AI is able to just take your whole mess and completely understand it because you can't even understand it. And we're working towards AGI and being as good as humans, if not better. But you know, even humans, you just like you know, you hear stories about associates joining funds and not knowing where things are and still struggling to navigate the file system with six months in. So I would say I think you know, people need to really start worrying about their data foundations and making sure that they have the right tooling in place and and vendors that they're willing to stay with for you know the the critical vendors, whether it's you know the the 73 strings of the world, your accounting system, because those are not likely, in my opinion, or too high impact to one get replaced and two be migrated from. Because even if you're able to build a portfolio monitoring system, I'm gonna give you the assumption that you're so smart that you were able to build that portfolio monitoring system. Migrating from your existing portfolio monitoring system to the new one is also another crazy topic. So yeah.
00:41:39 · Stuart Willson, S3, Joe Zein
[00:41:39] S0: You both spoke to the pace of change and the speed with which this technology is evolving. Where is portfolio monitoring in 18 months? What does the state of the art look like in 18 months? [00:41:52] S3: What's possible that's not possible today? I'm asking you to speculate. [00:42:02] S2: I mean, I think it's it's again, anyone who tells you they know what's gonna happen is either crazy or delusional. But I think it's not too far-fetched to imagine a world where there's an ability from those tools to ingest data that is being uploaded by portfolio companies or by teal teams or by the you know the portfolio monitoring teams that is also ingesting signals from the web and from the world, basically, and are able to produce the reports that you need automatically, and are able to mark your portfolio to market on a daily basis, and are that are able to surface insights for you without you having to do anything. So waking up to an inbox of like, hey, this thing happened in the world, you happen to have an investment in that place of the world, but your investment is has a high you know debt to equity ratio, you gotta be careful. Look into that specific lause of your covenant. Um I don't see, I mean, again, they've built the the building blocks, right? And I think the speed of of like uh you know, if LLMs keep developing at the pace they've they've been developing at, they'll be able to you know perfect the envelope around it so that uh you know it's really uh seamless where you're not doing anything anymore. People are able to do that.
00:43:26 · Yann Magnan, Joe Zein, Stuart Willson, S3
[00:43:26] S1: Yeah, but same thing. 18 months down the line, I see the same thing that uh you wake up in the morning, um, open up your laptop, and the popular monitoring system has run in in the background and provide you feedback, alerts, everything that you need to know. Same as uh if you had had the team working overnight and and getting everything ready. So all of that is on your laptop, and then it's on you to make the irrelevant decisions. And I think that's going to happen pretty fast, yes. [00:43:58] S2: But that's also using the assumption I I think to going on the assumption that you have that portfolio monitoring system, which which means that you have the data layer ready. Because assuming that you're not going to need a portfolio monitoring system, you're just going to upload all of your credit agreements and the internet and the LLM is going to figure it out, it's it's not something I see happening in the next 18 months. So I think I think it's it's if you have the foundation, this is what you can expect. If you don't have the foundation, you can't expect that. That's a great caveat. [00:44:32] S0: Yeah. That's a great setup for my last question, which is for funds that are exploring AI-powered portfolio monitoring tools for the first time, whether they are from 73 strings or they are bespoke or custom-built solutions, what's the one thing they should start doing right now to prepare? [00:44:54] S3: Call Leon? Yeah, I mean, I said that multiple times.