This isn't even the beginning. Transformer models show strong performance across a large variety of modes. The two most prominent so far are static image and hu

This isn't even the beginning. Transformer models show strong performance across a large variety of modes. The two most prominent so far are static image and human language generation. But it can also handle sensor input and motor signals. In a year or two anyone will be able to tape a mini PC to a drone with a stripped down 5.56 rifle and a GoPro and tell it to shoot anyone wearing the wrong football team's jersey. You might need to help it navigate at first but the image scanning and fire control will be fully automatic.