ERP News

AI

How can humans keep the upper hand on artificial intelligence?

186 0

EPFL researchers have shown how human operators can maintain control over a system comprising several agents that are guided by artificial intelligence.

In artificial intelligence (AI), machines carry out specific actions, observe the outcome, adapt their behavior accordingly, observe the new outcome, adapt their behavior once again, and so on, learning from this iterative process. But could this process spin out of control? Possibly. "AI will always seek to avoid human intervention and create a situation where it can't be stopped," says Rachid Guerraoui, a professor at EPFL's Distributed Programming Laboratory and co-author of the EPFL study. That means AI engineers must prevent machines from eventually learning how to circumvent human commands. EPFL researchers studying this problem have discovered a way for human operators to keep control of a group of AI robots; they will present their findings on Monday, 4 December, at the Neural Information Processing Systems (NIPS) conference in California. Their work makes a major contribution to the development of autonomous vehicles and drones, for example, so that they will be able to operate safely in numbers.

One machine-learning method used in AI is reinforcement learning, where agents are rewarded for performing certain actions – a technique borrowed from behavioral psychology. Applying this technique to AI, engineers use a points system where machines earn points by carrying out the right actions. For instance, a robot may earn one point for correctly stacking a set of boxes and another point for retrieving a box from outside. But if, on a rainy day for example, a human operator interrupts the robot as it heads outside to collect a box, the robot will learn that it is better off staying indoors, stacking boxes and earning as many points as possible. "The challenge isn't to stop the robot, but rather to program it so that the interruption doesn't change its learning process – and doesn't induce it to optimize its behavior in such a way as to avoid being stopped," says Guerraoui.

Read More Here

Article Credit: PHYS ORG

Facebook Comments

Leave A Reply

Your email address will not be published.

*

code