Researchers have developed a web-based application that shows people artificial-inspired paintings such as kittens and rhinos and records their reactions through the camera, which is a positive signal if people see them laughing.If people seem to Very confused, then it is a negative signal.
After gathering all of this feedback, the Google team re-trained the AI system and focused on the 'good' example, with the end result that artificial intelligence can create the paint better.
Although the experiment looks simple, the Google team points out that this ensures that AI is safer in the long term if AI is capable of adapting to human social behavior and can learn from human facial expressions and body language It's better to do this kind of behavior to make human happier.
Of course, the reverse is also true: Artificial Intelligence learns what behavior fears or disgusts humans, and MIT researchers are exploring the field, but Google is focusing on the positive side.
Google is not the only company to do that, and companies like Affectiva are exploring technologies that have artificial intelligence in mind and technologies that allow machines to better understand human facial expressions and social behaviors to make the machine more responsive to human needs, such as Affectiva Study how to make driverless cars understand that the driver is switching to autopilot mode.
Google's team pointed out: 'A human-driven artificial intelligence is not likely to be against the good of humanity.'