Google Says RT-2 Offers Glimpse of ‘Helpful Robot’ Future

Will Google’s RT-2 tool put us on the road to R2-D2?

That’s what the tech giant is hoping. The company on Friday (July 28) debuted its Robotics Transformer 2, or RT-2, which it called a first-of-its-kind vision-language-action (VLA) model designed to bring the world close to “helpful” robots.

“For decades, when people have imagined the distant future, they’ve almost always included a starring role for robots,” Vincent Vanhoucke, head of robotics for Google’s Deep Mind, wrote on the company’s blog. 

“Robots have been cast as dependable, helpful and even charming. Yet across those same decades, the technology has remained elusive, stuck in the imagined realm of science fiction.”

RT-2 can help make that fiction a reality, the company said. Trained on text and images from the web, the tool can directly output robotic actions, transferring knowledge from web data to inform robot behavior, similar to the way language models are trained on text from the web to learn general ideas and concepts.

“In other words, RT-2 can speak robot,” Vanhoucke wrote.

Unlike chatbots, he added, robots need to be grounded in the real world. It can’t just learn about apples; it needs to be able to understand what one looks like, distinguish it from things that look like apples and know how to pick one up. This has long required extensive training. 

“RT-2 removes that complexity and enables a single model to not only perform the complex reasoning seen in foundation models, but also output robot actions,” the blog post said. “Most importantly, it shows that with a small amount of robot training data, the system is able to transfer concepts embedded in its language and vision training data to direct robot actions — even for tasks it’s never been trained to do.”

While Google was making this announcement last week, The Wall Street Journal was profiling a Walmart facility where robots unpack trucks, determine where boxes should be sent, load and unload boxes from pallets. The retailer plans to have 100 such warehouses in the years ahead.

However, consumers aren’t totally sold on robots yet, as recent PYMNTS research has found. When it comes to robot delivery, 71% of consumers said they were not interested in robotics or automated systems bringing food.

“While the most common reason for that lack of interest is that consumers are worried about job loss and the lack of personal interaction, the second most common reason was that consumers do not trust these robots,” PYMNTS wrote recently. “Sixty-five percent of consumers who are not interested in robot delivery reported having concerns about reliability and order accuracy.”