Google is one of the companies that seems to champion the idea of AI but at the same time, they appear to have acknowledged some of the fears we might have, which is why Google-owned DeepMind has built a panic button of sorts that will be able to instantly interrupt AI and stop it from doing any harm, whether it be to themselves or to their human supervisor.
According to the researchers at DeepMind, “If such an agent is operating in real-time under human supervision, now and then it may be necessary for a human operator to press the big red button to prevent the agent from continuing a harmful sequence of actions — harmful either for the agent or for the environment — and lead the agent into a safer situation.”
However this isn’t a one-stop-shop for all AI. For example some algorithms such as Q-learning are found to be easy safely interruptible, while others might need modification. It is also unclear if this panic button can apply to all learning algorithms either, but we suppose it’s a start.