Interview Transcript

This is a snippet of the transcript, sign up to read more.

One of our internal theses on autonomy is that companies with the best access to data will ultimately be best positioned to achieve full autonomy over time. What's your take on that view? Is it correct, or would you disagree?

I completely agree. You took the words right out of my mouth. Data is crucial, especially now when we look at the big models, even in autonomous driving. Consider the foundational models from OpenAI, Google, and others. They have utilized almost all available data from the Internet or that they can purchase. However, they still struggle with tasks involving street scenes.

This is a snippet of the transcript, sign up to read more.

One of our internal theses on autonomy is that companies with the best access to data will ultimately be best positioned to achieve full autonomy over time. What's your take on that view? Is it correct, or would you disagree?

For instance, we experimented with Nvidia Cosmos, a model designed for robotics. When tested on series collected from, for example, Nexar cameras, it didn't perform well because it lacks sufficient data to train on those scenarios. It struggles to predict the next few frames and doesn't understand physics, creating physically impossible situations like cars passing through each other.

This is a snippet of the transcript, sign up to read more.

One of our internal theses on autonomy is that companies with the best access to data will ultimately be best positioned to achieve full autonomy over time. What's your take on that view? Is it correct, or would you disagree?

This indicates that whoever has access to valuable long-tail data on autonomous driving or driving in general can build or fine-tune a model like Cosmos better. Data is essential. With that in mind, who can win? Tesla is well-positioned because they collect a vast amount of data and have a smart way of collecting relevant data through their shadow mode in cars.

This is a snippet of the transcript, sign up to read more.

Sign up to test our content quality with a free sample of 50+ interviews