How Google AI Will Make Robots Smarter and What That Means for You

Your household chores may soon get easier as Google plans to boost robots' IQ with artificial intelligence (AI). 

The new computer technique, called Robotics Transfomer 2 (RT-2) could help train robots to understand tasks like throwing out trash. It's part of a growing number of methods to make robots smarter using AI that could transform everything from everyday home tasks to medical prosthetics. 

"By incorporating robotic solutions with AI, or more specifically machine learning algorithms, we enable them to learn as they perform repetitive actions and to make decisions about their next process autonomously and in real-time," Michael Nizich, a professor of computer science at New York Institute of Technology, told Lifewire in an email interview.

RT-2 Makes Smarter Trash Bots

In its blog post, Google announced RT-2, the AI model trained on vast amounts of information and images from the internet, empowering it to convert data into commands for robots.

Even though picking up trash might seem straightforward to humans, teaching a robot to do it involves comprehending a sequence of tasks. The robot needs to grasp the concept of identifying trash items, followed by knowing how to pick them up and dispose of them properly. Instead of explicitly programming these specific tasks, RT-2 lets the robot leverage web-acquired knowledge to understand and execute the job, even if it has not received training on the particular steps.

"It even has an idea of how to throw away the trash, even though it's never been trained to take that action," Google wrote in the blog post. "And think about the abstract nature of trash—what was a bag of chips or a banana peel becomes trash after you eat them. RT-2 is able to make sense of that from its vision-language training data and do the job."

According to Google, the latest model significantly improved the robots' performance in unfamiliar situations, nearly doubling their capabilities compared to the previous version. Google also claims the new iteration can employ basic reasoning skills to respond to user commands.

The kind of AI in RT-2 enables robots to make decisions, Elad Inbar, the CEO of robotics company RobotLAB, said via email. "Without that, robots aren't any smarter than simple machines like windshield wipers or toaster ovens," he added. "Artificial intelligence is basically a code written for that specific robot, processing input from sensors and coming up with a course of action based on that input. Without the decision-making process, robots are just automatons."

Autonomous cars are one example of how AI is used in robotics. The full self-driving vehicle uses artificial intelligence to collect and process gigs of information per minute from an array of sensors, cameras, and maps and produce the best course of action given all this information, Inbar said. 

"Whether we accelerate, brake, turn, or switch lanes depends on input from all sources, and then a probabilistic engine comes up with the most appropriate action to take based on all the input provided," he added.

Boom Times for AI Robots

Recent advances in AI fields, such as object recognition and natural language processing (NLP), are helping robots, Nizich said. Object recognition lets robots quickly identify objects and their relevance to their assigned tasks. NLP advancements empower them to have better-spoken interactions with their human counterparts. 

"The true acceptance of robots by humans will rely on true two-way communication between humans and their robotic assistants, which may be the only way that trust will ever be established in the systems by humans," he added. 

In the future, you might not need a keyboard to control robots. Nizich predicted that using robots with AI large language models (LLMs) like GPT-4 and brain interface technologies will allow the human control of commercial robots using just your thoughts. 

"The medical applications of this alone are enormous in areas like prosthetics, cancer treatments, and physical rehabilitation," he added. 

Humans might one day not even be needed when AI takes control of robots, Inbar said. 

"I see more and more products that are becoming autonomous and can make decisions based on their own judgment, to the extent that we won't even need to program robots anymore," he added. "We can just provide generic guidance, and the robot will figure out the rest."