The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
00:19:52 · Harry Stebbings, Adarsh Hiremath
[00:19:52] S1: Why then does so many people tell me, including Jonathan at Grok, that actually synthetic data is often more high quality? It doesn't involve the drags of the Internet like Reddit, in a lot of cases being included. And actually, you'll see this exponential increase in model performance due to actually mostly using high quality synthetic data, not low quality human data. Why is that wrong? [00:20:16] S0: So the first thing is that it's not zero sum. Even in a world where human data is super important for the next generation of models, it doesn't mean that synthetic data won't also be important. So synthetic data will certainly be a part of the equation. But in a lot of ways, the bottleneck to unlocking and unleashing the next level of intelligence will be expert humans, which brings me back to the question and phrase that you used, low quality human data. Low quality human data certainly won't push the models to be better at anything. High quality human data will. Again, that's a talent assessment problem. The biggest lever on data quality for creating these post training sets, for example, is finding the right people, which, again, is is really, really hard to do. [00:20:56] S1: In terms of, like, compute and algorithms, how do you think about where we're at today in terms of them being a bottleneck? We mentioned data is a bottleneck. Are compute and algorithms too? How do you think about that?
00:20:56 · Harry Stebbings, Adarsh Hiremath
[00:20:56] S1: In terms of, like, compute and algorithms, how do you think about where we're at today in terms of them being a bottleneck? We mentioned data is a bottleneck. Are compute and algorithms too? How do you think about that? [00:21:07] S0: All of them are pieces of the same puzzle, but the era that we're entering requires really, really expert humans to make models better at very, very specific use cases. [00:21:18] S1: How long will that be the case for? [00:21:20] S0: For a very, very long time. There's this huge long tail of tasks that models can't do. [00:21:27] S1: Like what? Just help me understand. You're teaching me.
00:21:27 · Harry Stebbings, Adarsh Hiremath
[00:21:27] S1: Like what? Just help me understand. You're teaching me. [00:21:30] S0: Well, maybe we could take a step back. If we reach the point a couple of year... A hundred years from now where the models are able to do every single job and humans no longer have any work to do, society is gonna look really, really different. We're all going to be living on a UBI. We're all going to be playing video games all day, whatever it may be. But until that point comes, there's going to be a whole set of tasks that the models cannot do, whether they be specific economically valuable tasks, like maybe the job that a consultant could do or maybe a specific category of engineering or even like more niche things. Right? Maybe it's making the model better at like some specific hobby. And we're always gonna need to fill in the gaps, particularly in that long tail. And Harry, the other thing I'll say is I think people are really, really in this mindset of, like, this, like, unidec directional relationship between humans and AI. Right? Where I can't do something, I give it to the AI, the AI takes it to completion. But I think the more realistic breakdown is the AI for a specific use case might be able to get us 60, 70, 80% of the way there. But for that remaining 40, 30, 20%, you're going to need a human to be able to to take you all the way there. And the reality of the situation is finding that human will become harder and more valuable to do.
00:22:50 · Harry Stebbings, Adarsh Hiremath
[00:22:50] S1: If you get further and further up the spectrum of we're able to 70, 75, 80, 85, 90? Do you not need fewer and fewer humans because the frequency is much less as you move closer and closer to perfection? [00:23:04] S0: That's a great question. And I think that begs the question of what do labor markets look like later on. I think the key thing is that the market will move towards specialty and sophistication, meaning the types of work that we see fifty years from now will be more specialized and oftentimes will require people with kind of like a higher level of sophistication in that specific thing. [00:23:27] S1: Can I ask you, when you sell to clients, what is the moment where they're like, wow, shit, we've got to use my core? [00:23:34] S0: When we're able to find exceptional people at the cost of software hundreds of times over. [00:23:39] S1: But when you're in that sales cycle with them today, when are they going, yeah, we've gotta sign up? Is it when they see the AI interviewer? Is it when you show them the price? Is it when they meet a candidate? Like, when when is that wow moment for them? [00:23:54] S0: It's usually when the first couple candidates start working with them. [00:23:57] S1: What what is that buying process? They buy one at a time? Is it on a per talent basis? Is it on a timeline basis? How does how does the deal with work?
00:23:57 · Harry Stebbings, Adarsh Hiremath
[00:23:57] S1: What what is that buying process? They buy one at a time? Is it on a per talent basis? Is it on a timeline basis? How does how does the deal with work? [00:24:05] S0: One interesting thing about Merkur, we don't have a sales team. There's not a single person who works, on sales at Merkur outside of, you know, the founders. These days, what we're seeing is mostly customer inbound. Folks have heard great things about Merkur from other people who have hired through Merkur and then reach out to us, and then we go from there. So right now, it's more of a a bandwidth thing than than any, like, tactical or coordinated sales motion, I would say. [00:24:32] S1: What percentage of hires is end to end done by software versus has human in the loop? [00:24:38] S0: On our end, the entire process is automated. This is everything from a candidate hearing about Merkor and going onto the Merkor platform via job listing, us pulling in their resume, their salary expectations, and whatnot, administering a personalized interview based on both their background and the role, allowing them to get paid for their work. That entire process is automated. [00:25:02] S1: What does a take look like on a per candidate basis?
00:25:02 · Harry Stebbings, Adarsh Hiremath
[00:25:02] S1: What does a take look like on a per candidate basis? [00:25:06] S0: It all comes back to quality. You know, I briefly talked about Uber and just going back to that example. Right? When I get into an Uber, there isn't that much of a difference between the 4.8 star driver and the 4.9 star driver, because the unit of work is not exponential. But with something like Merkur, there's a huge difference between the top point 1% and then the eightieth percentile person. So usually for customers, it's not a question of price. It's a question of quality. If we're able to find those point 1% people reliably at the cost of software, what what we take is often a second thought. [00:25:43] S1: And so I'm sorry. I'm... What what does that take then? Is it like a standardized take? Is it like on a case by case basis? What does that look like? [00:25:51] S0: It's on a case by case basis. For some customers, it can be, you know, over 30%. For some, it can be less. [00:25:56] S1: Over 30. That... That's great. Amazing. Well done. When you look at candidate completion rates, how much of that is India versus rest of world today? I know you specialize in finding amazing talent in India specifically.