Research shows AI far from an autonomous life of its own
Humanoid robots operated by artificial intelligence (AI) have claimed that they will one day be able to rule the world more efficiently than humans. It was during the World Summit on AI for Social Good, organized by the ITU, the UN’s specialized technology agency, in July. However, the machines admitted a challenge to be overcome: mastering human emotions. A warning for human beings to beware of technologies and their advances.
The episode caused a great shock in the audience and was shared with a certain alarmism and sensationalism about AI predictions. Marcos Barretto, professor at the Polytechnic School of USP, co-founder of StartADAM and angel investor of Poli Angels, says that the answers given by the robots still depend on training done by humans, that is, they do not have autonomy.
Demystification
The professor emphasizes the importance of recognizing any conversation between humanoids and the people present there as something programmed by humans. “Someone programmed and trained the robots to say something, so it’s a bit like theater in that respect, there’s a script, which was rehearsed for the machines to respond”, explains Barretto.
Thus, the idea of a certain autonomy of the humanoid in the statements does not materialize, since there are still limitations that depend on human action. In addition, the professor considers that there were also other devices that were optimistic about the contribution of technology to society.
emotional challenge
Human emotions, as well as the way they are transmitted in the voice and gestures of individuals, represent central barriers between humanoids and their models. In Barretto’s view, there is great difficulty in finding algorithms capable of understanding feelings and making technological devices behave accordingly.
“For example, you are watching a movie and a robot vacuum cleaner decides to pass by at that moment. So, the design of the advances of these technologies starts from the understanding of the social moment in which you are living through the machine”, exemplifies the professor.
Other contexts for the application of this technology were also mentioned, such as a possible understanding of situations in traffic, in which the driver is inattentive and the machine develops some device to understand this social moment.
ChatGPT
The responses generated by the robots at the UN meeting also used ChatGPT for interactions, according to a language model that is based both on information available in books and on knowledge from the responses of humans themselves. Thus, Barretto points out that despite the large volume of information acquired by machines compared to the human reality of learning, robots are far from the stage of their own life.