Robots need a lot of data
It takes a huge amount of data to train robots to live and work autonomously alongside us. Even with simulations and other ways to create training data, it is unlikely that robots will ever start working on a basic model that controls the entire system.
It’s still unclear how complex tasks machines can master using AI alone. According to Brondmo, thousands or even millions of robots would need to be trained in the real world to collect enough data to end-to-end train models that can perform more than just well-defined tasks.
It will take a long time for both artificial intelligence and canada number data traditional programming to create useful robots that, for example, clear tables in a restaurant or make beds in a hotel. In other words, don't expect robots to get out of control and start doing things they weren't programmed to do anytime soon.
Brondmo believes that robots shouldn’t just imitate us. He learned this from a meeting with the technical leaders of Everyday Robots. The participants sat around a negotiating table and had a lively discussion about whether robots should have legs or wheels. One side offered a good argument: the places we live and work in are designed for us, and we have legs. So maybe robots should have them too.
Half an hour later, Vincent Duro, the oldest engineer in the room, said, “I think if I can get somewhere, so can the robots.” Vincent was sitting in a wheelchair. The room fell silent. The discussion was over.