Real Alpha
Watch mode

Sabrina Halper Show · 5 min · Oct 15, 2026

Training data is a strategic asset that should not be exported

Ali Ansari

Founder & CEO, Micro1

Founded micro1, pivoting it from AI recruiting to expert training data, growing it past $1B ARR.

  • Sabrina HalperHost · Host, Sabrina Halper Show

    Has worked as an investor at HOF Capital and founded SHS Media.

About this segment

Ansari argues that structured training data carries valuable institutional knowledge and gives models their intelligence. He warns that selling it to geopolitical rivals could help them catch up quickly.

From “AI Will Always Need New Data to Stay Aligned With Humans | Micro1’s Ali Ansari”, Oct 15, 2026. Watch the full episode

“what is the thing that actually gives models their, like, IQ points, their, like, real intelligence, it's the data piece, right?”
Ali Ansariat 50:15 · Sabrina Halper Show50:15

Takeaways

  1. Data, compute and algorithms drive intelligent systems.
  2. Ansari sees data as the source of models’ IQ points.
  3. Structured training data can transfer institutional know-how.
Watch 4:41 from 49:24

In this episode

Full episode
  1. 22:36Human expertise must keep updating AI models5:45
  2. 35:18Robotics is held back by a much larger data problem4:40

  3. 49:24Training data is a strategic asset that should not be exported4:41

  4. 5 more moments in the full episode
Up next · Core MemoryRealbotix prioritizes autonomy and useful tasks over walking6:23

Transcript

Transcribed by machine (Whisper) from the episode’s audio, so a word may be off here and there.

  1. 49:30

    Yeah. So yeah, this is a really important topic to us at Micro One. So just to take a step back, the things that allow intelligent systems to emerge and improve, which can be models, agents, hardware embodiments of these models, whatever, just intelligence systems broadly, is data, compute, and algorithms. Compute is incredibly important. It's how you run trainings, how you serve the models. It's a huge bottleneck. And we're working on it.

  2. 50:04

    Yeah, and I'm working on it. There's a ton of investments there. Algorithms are also very important. It's what allows for models to improve even faster. and there's a ton of researchers iterating on new architectures, you know, that's sort of there as well. But if you think about, like, what is the thing that actually gives models their, like, IQ points, their, like, real intelligence, it's the data piece, right?

  3. 50:30

    And if you think about where are we ahead compared to, you know, countries that we're in a race with, we are very ahead on the intelligence piece. We're not ahead on hardware. Um, uh, and, you know, algorithms, I think it's, we're, we're sort of ahead, but, but there's, you know, it sort of gets commoditized a lot faster. So, so, so the key thing that allows the models to have the IQ points that they have and the IQ points that we keep sort of increasing is, is the data.

  4. 50:58

    And, and, and, and if you, and, and so if that's the area that we're ahead and we believe it's the key thing that matters most, then it's the area that we should not be exporting to, to, to our adversaries. And if you think about like what is an RL environment, it's basically a package up, it's a very compact, very structured, you know, sort of object, right? like digital object that, that, that includes a ton of intelligence within it. And, and if you, if you have a bunch of American intelligence and, and company operational data to seed these environments and, and you're, you're shipping it and you're exporting it to, to, to nations that are adversarial to us, what you're doing is you're essentially just exporting intelligence in the U.S.

  5. 51:51

    at a very large scale in a very structured way and you're allowing our competitors to catch up very quickly and what I would argue is that the way that, you know, the sort of adversaries are catching up to us is this. Like, this is the main thing, which is you are other data labs in the space are selling large-scale off-the-shelf data sets and environments to these other labs. And there's also distilling that's happening.

  6. 52:22

    But I would argue that it's actually the former that is even potentially more important that is resulting in these mobile companies catching up. And to us, that's not the right thing to do. China has a lot of people and a lot of amazing companies, too. So when we're selling training data, do you think it's that, like, our data is actually something that is precious in a sense that, like, they can create?

  7. 52:46

    Or it's just, like, we already have these giant companies doing it, so we're just helping them expedite frontier intelligence? I think it's exactly both of those things, which is, you know, I believe we have the best institutions in the world. uh not only academic but also commercial institutions these institutions innovate a ton the data that comes from the institutions that innovate most is uh is the most valuable data so so yes american data is the most valuable data um and uh and there are companies you know data companies that do this at a very large scale that are the ones that are making sure we could keep pushing the frontier. And so that large scale operation itself is also one that we shouldn't allow our adversaries to have in a sort of easy way. So it's both, yeah. Also, one thing we've

  8. 53:38

    heard is like Chinese models obviously have like more Chinese training data, right? So it's like as they're open sourced and used in the US, there are certain things we might not even be aware of that they're just like not as good at, which this, I guess, would totally solve. So it makes them a better competitor okay so I have heard rumblings and I don't expect you to confirm this