Snorkel AI

So the bottleneck has shifted from acquiring just a sheer volume of data to acquiring domain-specific behavioral data.

And that brings us to the staggering numbers coming out of Snorkel AI.

Yeah, let's talk about Snorkel.

They just closed a $350 million Series E funding round, which propels them to a $3.5 billion valuation.

OpenAI dice que no reanudará el proceso hasta contar con nuevas salvaguardas y advierte públicamente que probablemente tendrá que volver a pausarlo.

En el frente de financiamiento, Snorkel AI acaba de cerrar una ronda serie E de trescientos cincuenta millones de dólares con una valuación de tres punto cinco mil millones de dólares.

Snorkel construyó la infraestructura de etiquetado de datos de la que dependen otras compañías de inteligencia artificial.

Así que cuando los laboratorios gastan más, Snorkel crece con ellos.

it's a great question.

So the type of companies like Snorkel and Patronus, you know, I'm a big fan of, you know, full disclosure, I'm actually angel investor in Patronus.

And so, you know, have a lot of respect for those companies and rooting for them.

We do something that's a little bit different.

Yeah, actually the, um, the best model, like still Fable is only cracking 6,025 Tasker.

And the reason why we pick like Kickadd, like, um, that was a benchmark we made in collaboration with, uh, Snorkel.

And the idea behind it is that you can measure like, um, if like real trajectory, um, are gonna, um, gonna like basically end with like your end result and you can basically like measure it very programmatically using like keycard as code.

So it's like a very problem, like very hard problem to crack when you're like interacting with these different components that you need to drag around on a canvas.

To the funding rounds.

Snorkel AI raised $350 million in a Series E at a $3.5 billion valuation, nearly tripling its mark from 17 months ago.

Insight, Partners, and S32 co-led the round, with participation from Third Point, March Capital, Bloomberg, and existing investor Addition.

The seven-year-old company builds training data sets in simulated environments for frontier AI labs and large enterprises.

So that's really nice.

And then this week I had some training on an app called Snorkel.

And people are surprisingly surprised about the things that I'm doing with this AI and just getting them to think about thinking and not really about AI.

So a lot of my trainings are all about how do we add that productive struggle? How do we make sure that we're still developing those foundational skills? How do we make sure that kids are still learning? And most importantly, how do we know that the kids are learning? And so it's been real fun.