{"id":585256,"date":"2023-08-01T19:39:27","date_gmt":"2023-08-01T16:39:27","guid":{"rendered":"https:\/\/en.buradabiliyorum.com\/deepminds-new-ai-controls-robotic-tasks-without-specific-training\/"},"modified":"2023-08-01T19:39:27","modified_gmt":"2023-08-01T16:39:27","slug":"deepminds-new-ai-controls-robotic-tasks-without-specific-training","status":"publish","type":"post","link":"https:\/\/buradabiliyorum.com\/en\/deepminds-new-ai-controls-robotic-tasks-without-specific-training\/","title":{"rendered":"#DeepMind’s new AI controls robotic tasks without specific training"},"content":{"rendered":"
\n Google DeepMind has a new AI model that can direct robotic tasks it was never trained to perform.<\/p>\n

Named RT-2, the model learns from web and robotics data. It then turns this information into simple instructions for machines.<\/p>\n

In tests, the model was asked to take actions never seen in the robotic data, such as placing oranges in a matching bowl. To follow these commands, the system had to translate knowledge from web-based data. <\/span>According to DeepMind, the model had a 62% success for these operations \u2014 double that of its predecessor, RT-1.<\/p>\n

\u201cJust like language models are trained on text from the web to learn general<\/a> ideas and concepts, RT-2 transfers knowledge from web data to inform robot behaviour,\u201d said Vincent Vanhoucke, head of robotics at DeepMind. \u201c<\/span>In other words, RT-2 can speak robot.\u201d<\/p>\n

\"The
<\/a>RT-2 was tested on various robotic skills that weren\u2019t present in the robotics data. Credit: Google DeepMind<\/figcaption>