The Ledger, every show

“The programmatic analysis of early stage companies, I think, is a fool's errand, and I have seen some of the best data in the world of AngelList. I've talked to all sorts of people experimenting with that data when I was there and since then, and it doesn't re…”

“You need proprietary data to build a model that generates results that are far better than what someone else can generate by throwing commodity data into, you know, an openly available model.”

“In SaaS, you're competing on features and integrations, and they're easy to copy, and it's just a rat race. And in AI-first companies, you can get this runaway advantage very quickly because the system compounds in value itself, like without it, without any ef…”

“You tend to start at 40 or 50% gross margin, but then you get to 95, 98%, we're seeing in some companies we work for gross margin, and then you just stay there. As opposed to SaaS companies, we're constantly having to build another feature or deliver a service…”

“There's more research papers, AI research papers that is coming out of Europe than anywhere else, but European AI entrepreneurs get one-tenth funding of US entrepreneurs.”

“And we met, and we just had this mind meld around, like, this next era of computing, and that is, all systems, all software will become intelligent software.”

“And what's really gonna happen in that era, which we're sort of starting now, is You get an ensemble of AI techniques, so you might have like a Bayesian system and a neural network and a bunch of other stuff that work in concert together to understand a system…”

“You could actually think of what we do as value investing in a way, because what we do is we price the risk that a company can actually generate this really valuable prediction. And when we meet a company, they're probably at 30 to 50% accuracy on a model, whi…”

“And so, our investors don't have to hold a belief that we're extraordinary pickers that can predictably find companies in their very, very long tail of the power law distribution. We hope we do that, but they don't have to believe that up front if your fund si…”

“A lot of the, all the companies in the market today for search as a service are AltaVista, and constructor is Google.”

“And some of those customers will ask you to do an on-prem private deployment, will ask you to not use their data to train anything that you do, and you need to reject those customers. Because you are not going to build a company that has any sort of moat aroun…”

“If you're a company that's not an AI first company that started sort of before this AI era, you can't really build any predictive models of consequence because you don't have any of that customer data and you have to sort of go back to your customers, renegoti…”

“Data moats are indeed real, and we evaluate those data moats very carefully at the diligence stage, and we look at things like accessibility, fungibility, dimensionality, breadth, and perishability.”

“And you know, if that number, which is the potential market cap of the company discounted to today is bigger than the size of our fund, then that's the starting point for thinking about making investment.”

“I'm really hesitant to build work relationships this way because it can discriminate against people in all these little ways. So I think the only thing to try to answer Andy's question, the only thing I've found is, you know, just go on long walks with people …”

“We've stayed away from a lot of applications of current AI technologies to healthcare, because when it goes wrong, it's really catastrophic, and it will go wrong. But we've moved towards a lot of applications to it in industry, like, again, in manufacturing an…”

“There's more AI research coming out of Europe than anywhere in the world. More than the US, more than China.”

“I used to think that geographic advantage in sourcing is like one of the most unsustainable things, because anyone can just get on a plane and be somewhere, but it just constantly surprises me that just being in these centers and being able to have conversatio…”

“You want to work in industries where the payoff is really high. So I think of like convex and concave payoffs, where if you get it right, it's really valuable to you. But if you get it wrong, it doesn't really matter. So again, going back to medical use cases,…”

“The reality is, as much as we're excited about this era of AI creating new knowledge and ensembles of models, understanding complex systems, for the most part, we're pretty far away from that, and the reality is a lot of AIs are, like, fairly simple statistica…”

“I think the interesting thing, and this has been something that I'm a bit of a stickler for, is our memos start as a blank page, and we try not to use templates, because I think templates form your thinking, and so you start with a blank page, and you just sta…”

“I sort of say something internally, which is not often the most helpful thing to say, which is as soon as you're competing, you're losing.”

“It pains me to think that we're going into an era where no one cooks their own food, but we are, and delivery and various other technologies are getting so good that you click a button and you get food in 10 minutes, and people love that, of course.”

“I think also people think about data volume rather than the dimensionality of data or whatnot, and often you solve really important problems with really small amounts of data. And so, I think people think about volume, and data is the new oil, and all this sor…”

“I also see every day people breaking off from these companies and just applying their knowledge to all sorts of other industries day-to-day and starting their own companies, and so Talent is like not that concentrated, actually, and there are all sorts of grea…”

“From the very beginning of our fund, we just said we're not going to invest in anything that's horizontal.”

“What we don't look for is traction, right? Like by the time a company has revenue and whatever else, it's sort of, One, if you think of what my job is, it's to find things that other people can't find that are valuable, and it's pretty obvious at the point whe…”

“And at AngelList, we found companies like Ubers, Imogen, and Opendoor before anyone else, or before many others.”

“AI first companies, on the other hand, they focus on building predictive models, not product features, getting quantitative feedback, not qualitative feedback, and from machine agents, from how the model's just running itself rather than human customers, and t…”

“A system of record gets data in a structured form as it's manually entered by a human sort of after the fact in like instruction fields, so like a CRM or an ERP or something like that. A system of intelligence gets data usually in unstructured form, and it's a…”

“AI first companies tend to have a little bit more of a shared benefit model or a percentage of the increase in revenue model.”

“The one thing that you take away from that, or I took away from that, that I apply to venture is not like, this is how you hire your first salesperson, or this is how you like integrate with this platform, because all that knowledge changes, best practice is a…”

“That was a really interesting time, 2013, 14, because that's when this neural network revolution started, and that is, we were finally at the point where we'd had some research breakthroughs into how neural networks work. We had enough data to feed these, like…”

“Once you get to like a specific industry problem, like Is the bottle cap too big for the bottle on this production line? Like, did the plastic extrusion process not work properly? You can't sort of use off-the-shelf models to do that, and so, what you need …”

“And so working out what that minimum, we call it the minimum algorithmic performance, but essentially like what accuracy a customer needs for that to be valuable to them is one thing. And then how far away is the company from achieving that?”

“And so at Zeta, if you meet a company, you have to decide whether to have them meet someone else on the team or pass on the opportunity after the first meeting.”

“Firstly, do I even have the activation energy to write this memo? Because if I don't have the activation energy to write the memo, am I going to want to turn up to board meetings on like a six hour flight away every month? Am I going to want to get on the phon…”

“And the unintuitive thing about power law distributions is that your average return goes up as your points of exposure go up.”

“What I'm excited by there is not necessarily like the next best data source, because they tend to be very perishable. I'm excited about different tools for those hedge funds that are trying to make predictions in complex systems.”

“The state of tooling for a data scientist in, so what, 2019 is what it was for a software engineer in like the late eighties, early nineties, There's still so much that's so hard about building these things that could be made a lot easier with the right tools”

“They're not really structured, and their businesses are so big such that it's not really worth it for them to go in and solve problems in a lot of vertical markets, so markets where the total opportunity is a couple hundred million dollars. This just doesn't w…”

“I tend to switch off in pitches where the presenters sort of use like a word cloud of popular terms in their pitch, and it's also a pretty strong sign to me that they haven't come to a fundamental understanding from something new. Like, to explain something bo…”

“We lay claim, which sounds like a funny claim to lay today as being the first fund completely focused on intelligent systems.”

“What we're not spending time on are things where AI, well, we don't spend any time in the consumer stuff because a lot of that data is owned by bigger companies, and it's very hard to get a data advantage there, and we don't really spend time on things that ha…”

“To make us only invest in vertically focused applications, because that's where you can really get ahead of everyone else by focusing on tuning a model for a very specific purpose, getting data to train a model for a very specific purpose.”

“The tools that data scientists have available today are like what software developers had in the eighties.”

“Now, Algolia, Elastic, like, all these companies do that. They make it very easy to deploy a very fast search box on your website, but the thing is, because they guarantee you that they're not going to share any data in any way with anyone they can't really im…”

“Looking really far, you know, relative to Silicon Valley mindset, really far afield into some of those disciplines to hire people has been immensely productive for companies that we work for.”