AI's Unsung Hero: Data Labeling and Expert Evals

**Manu Sharma** (0:00)
Somewhere around the GPT-3, Dali, the kind of the first phases of models where we were starting to see, something fundamentally was changing. Supervised learning was taking a backseat, and rather unsupervised learning was starting to work. Around the ChatGPT moment, we started to see RLHF emerge, where it is rather tedious to ask people to write the essays or solve some problems from scratch, but we can capture preferences very easily from humans and experts across different fields. Now we are in 2025 in a regime where reinforcement learning has came back and is a new technical vector that all of the AI labs are scaling. My best way to describe it is like meta-learning. Instead of telling a computer what is good or bad, the experts are essentially trying to teach these algorithms how to assess what is good or bad.
It's not simply like getting the answer right, it's how great the answer is.

**SPEAKER_2** (1:01)
Thanks for listening to the a16z AI podcast. In this episode, we dig into one of the unsung heroes of the AI industry, data labeling and evaluation. Now you've likely heard about Meta's big investment in Scale AI. But before that news, it was still an incredibly important piece of the model training pipeline that largely flew under the radars of non-practitioners. So we brought in Labelbox co-founder and CEO Manu Sharma. We sat down with a16z Infra partner Matt Bornstein to explain the foundation and evolution of data labeling and how his company has been able to ride that way from computer vision to reasoning models to, more recently, helping power advances in state-of-the-art voice models. As Manu explains in detail, there was a seismic shift over the past several years as the value moved from labeling pre-training data to evaluating outputs of the reinforcement learning phase, signaling a shift in model capabilities, architectures and applications, as well as an even greater need for human experts to help models perform across more complex modalities and with more demanding users. It's a great introduction to this space, and a great example of being able to ride the wave as a founder and startup, and you'll hear it all after these disclosures. As a reminder, please note that the content here is for informational purposes only, should not be taken as legal, business, tax or investment advice, or be used to evaluate any investment or security, and is not directed at any investors or potential investors in any a16z fund. For more details, please see a16z.com/disclosures.

**Manu Sharma** (2:34)
2014 to 16 or 18 was some really, really interesting times where we were starting to, for the first time, see computer vision algorithm starting to work. And I was working at a couple of technologies in space industry. I was looking at building technologies at planet labs, where we scanned the Earth every day with all 300 or 400 satellites orbiting the Earth. And there was just so much vast data, where it was basically the obvious thing to do was to use the machine learning algorithms to extract insights from the data and power the geospatial industry with those insights. And it was around that time where I felt the need for the data is so essential to developing these models that we could build something here. We could build a product here and so forth.
And I think that really led to building Labelbox. We launched it on Reddit, among all things. And our initial prototype got so popular that in the weeks after the launch, we just started subscribing customers. And our customers were across the sector. I think they were health care customers or robotics or geospatial and insurance. And it was just such a kind of exciting momentum to see in that we kind of rolled into building a company around that, essentially.

**Matt Bornstein** (4:05)
So these were the, I would call it early days of traditional machine learning really starting to take off.

**Manu Sharma** (4:11)
That's right. Exactly.

**Matt Bornstein** (4:12)
And you were kind of there helping it happen.

**Manu Sharma** (4:15)
That's right. Exactly. So the self-driving cars were, companies were starting to pop up. And then there were kind of this kind of few big companies that had just vast amount of data and they were applying these computer vision algorithms to kind of see what products and capabilities they could build. And so it was sort of very early innings. However, it was evident for a group of people who had been in the industry for a while. So I remember like in 2010 or 2012, and I was kind of in academic programs, my neural network kind of work would be like three layers of neural networks and maybe 10 neurons, you can count them. And they were on Matlab, Simulink, and you kind of use that to test these things and train these networks. And so it was already kind of a vast amount of progress from 2008, let's say to 2016