3 ms·

Scale AI's secret sauce is in labeling lidar point clouds, which is really a necessity for the self-driving car industry. Sure, they can label other stuff as w

Scale AI's secret sauce is in labeling lidar point clouds, which is really a necessity for the self-driving car industry. Sure, they can label other stuff as well, but it's not hard and there are other companies that can do it (not quite as well, but they could easily catch up). What I'm trying to say is Scale AI is pretty much the only company offering a solution for an absolutely hair-on-fire problem for self-driving car companies - the introduction was unnecessary (and many of their customers are Argo, Waymo, Zoox, and other companies not affiliated with Y Combinator). Assuming the self-driving bubble deflates (it will) then Scale AI will suddenly find itself a commodity business as labeling images is not so hard that it can't be easily copied. It will be a race to the bottom. Unless there is another sudden surge of demand for labeling lidar point clouds.

m_ke 7y ago
Yeah I know what they do and know a few other people working on data annotation tools for autonomous vehicles. All of the data annotation startups pivoted to that space because of the money involved. I just used them as an example because they were the first recent YC company that came to mind that's working on enterprise tools and has a landing page full of logos.
- OnlineGladiator 7y ago
  Fair point. Sorry if I came across as argumentative - I just thought it wasn't the best example. Regarding your overall point of YC having a network effect - not only is it true, it's part of their thesis for scaling up YC as much as possible. YC made a big push when Sam Altman became president (he's now heading up OpenAI) to try to grow YC as much as possible. This was about 5 years ago. The real question is how will the network effects weather the storm of a recession, which is untested (YC really started right around the last recession). Are they inside their own echo chamber? Are they actually supporting self-sustaining businesses? Will the network become big enough that it can sustain many of its businesses? I'm curious to see how the tech startup scene shakes out with the next recession. Let's just say I'm not considered an optimist.
coderunner 7y ago
Does Scale AI use trained human annotators to label point clouds for customers or do they have a trained model that they use to quickly and automatically label point clouds for customers?
- tziki 7y ago
  Using an ML model to label point clouds for generic ML purposes sounds like a horrible idea.
- akelly 7y ago
  I met an ex-employee, they have 500 employees in the Philippines doing annotation. They built some software to help annotate point clouds but it seems like mostly a contractor handling logistics that the tech companies would rather not deal with, and they became a unicorn because AI.
  - coderunner 7y ago
    Oh okay, thanks. I wasn't sure if they were selling annotations that their trained model can do quickly or selling manual labeling services that their customers can use to train their own models. Sounds like the latter which isn't as technically impressive.

m_ke 7y ago

Yeah I know what they do and know a few other people working on data annotation tools for autonomous vehicles. All of the data annotation startups pivoted to that space because of the money involved. I just used them as an example because they were the first recent YC company that came to mind that's working on enterprise tools and has a landing page full of logos.

OnlineGladiator 7y ago
Fair point. Sorry if I came across as argumentative - I just thought it wasn't the best example. Regarding your overall point of YC having a network effect - not only is it true, it's part of their thesis for scaling up YC as much as possible. YC made a big push when Sam Altman became president (he's now heading up OpenAI) to try to grow YC as much as possible. This was about 5 years ago. The real question is how will the network effects weather the storm of a recession, which is untested (YC really started right around the last recession). Are they inside their own echo chamber? Are they actually supporting self-sustaining businesses? Will the network become big enough that it can sustain many of its businesses? I'm curious to see how the tech startup scene shakes out with the next recession. Let's just say I'm not considered an optimist.

coderunner 7y ago

Does Scale AI use trained human annotators to label point clouds for customers or do they have a trained model that they use to quickly and automatically label point clouds for customers?

tziki 7y ago
Using an ML model to label point clouds for generic ML purposes sounds like a horrible idea.
akelly 7y ago
I met an ex-employee, they have 500 employees in the Philippines doing annotation. They built some software to help annotate point clouds but it seems like mostly a contractor handling logistics that the tech companies would rather not deal with, and they became a unicorn because AI.
- coderunner 7y ago
  Oh okay, thanks. I wasn't sure if they were selling annotations that their trained model can do quickly or selling manual labeling services that their customers can use to train their own models. Sounds like the latter which isn't as technically impressive.