Sorry but I don't believe what the title proclaims. And I don't even think the author does: "Thus, the large, high-quality dataset becomes a more complex challenge to solve in robotics, and deploying a fleet of robots in production is the only way to build a diverse dataset." -- So you acknowledge that we haven't even started gathering the data we need to train the model but compare it to GPT which was trained on the entire Internet, which, fortunately, already existed. So why exactly are we near?