Google Tests Robot Morality

Google has developed an AI model that can control robots, blending language, vision, and physical-data training — allowing them to do physical tasks and, allegedly, make decisions about their behavior to guard against doing something unsafe.

Why It Hits: As chatbots take over the mainstream, robotics is seen as the final frontier of AI. Several companies are racing for dominance — Google, OpenAI, Tesla, Anthropic, Figure, etc. — and whoever is the first to roll out a fully-autonomous robotic workforce may print an untold amount of money.

Behind the Code: Google’s “Gemini Robotics” is being touted as the cutting-edge system to get robots to accomplish any task… legally speaking.

• Powered by voice commands, robots equipped with the system can do things like fold paper, put on glasses, and pass items.
• The robots don’t need to be trained on specific tasks but instead learn to do things through understanding the general concept of what’s being asked.
• Additionally, the system is loaded with a “moral” benchmark called ASIMOV, which is named after the famous sci-fi author who came up with the “Three Laws of Robotics” (maybe you remember them from the I, Robot marketing).
• According to Wired, ASIMOV allows the robot to determine whether its actions could result in something dangerous by “presenting it with a multitude of different situations.” It can then, theoretically, stop or alter its behavior.

The Future: Google is already working with robotics firms like Agility Robotics, Boston Dynamics, and Enchanted Tools to test the software. Although Google robotics lead Carolina Parada says that the company doesn’t have any plans to commercialize the tech, Google is rarely in the business of keeping a moneymaker private (especially after playing catch-up with OpenAI and chatbots)… so expect there to be a big “yet” to be silently added to the end of Parada’s claim.

Google would love to create the Android of robotics.