Microsoft says it has used the natural language Al ChatGPT to control robots with simple text commands. The approach means people with no engineering experience will be able to instruct robots to carry out tasks.
Microsoft said the research was intended "to see if ChatGPT can think beyond text, and reason about the physical world to help with robotics tasks". Robots are typically controlled by software that has been written by humans and sets out precise instructions, or else by some form of network Al that can learn to carry out tasks based on large numbers of examples.
Neither Microsoft nor Open Al responded to a request for comment on this, but Microsoft said ChatGPT allows a user to monitor the process. ChatGPT isn't in direct, real-time control of the robot, but simply creates the code that controls it. Microsoft said ChatGPT output shouldn't be used directly to control a robot without "careful analysis", but experts warn that it is risky to even begin thinking about placing Al in control of physical machines.
Mark Coeckelbergh at the University of Vienna, Austria, says that regulations may be needed in order to set out where Al can be used and who is responsible should things go wrong. "It's a very dangerous thing to just say ‘let's give control of these computers to AI'," says Coeckelbergh. "The problem with contemporary artificial intelligence is that it's not transparent(透明的) to the user. That's a huge problem."
Others were doubtful about the ability of language models to program robots for complex tasks. Kathleen Richardson at De Montfort University Leicester, UK, says that Al language models are convincing mimics (模仿者), but that their real ability — including their adaptability to control robots — is often overblown. "I think Microsoft, and most people who design robots,overstate what they can and can't do," she says.