People often ask us, “What is the point of video game benchmarks, and how do they help solve real-world problems?”
Below this post, we have shared a video highlighting the progress we are making in our robotics research. This video shows the AXIOM digital brain approach, previously seen in the Gameworld benchmark, applied to robotics.
Benchmarks show that we can solve a critical problem: How models can learn quickly, starting from a simple goal, such as unpacking a fridge or scoring points in a game.
Robots often perform well on scripted tasks but can freeze when faced with new situations. Even something as simple as a box in the wrong place can halt progress. We aim to change that by giving machines a brain that, like our human brain, learns from experience, makes sense of uncertainty, and adapts in real time.
In the video clip, we show how a robot decide to find its own way to get to the table because it can’t reach its target location without going around the sofa. By contrast, many traditional robotics approaches can’t “think” about alternative approaches, unless they have had specific training.