Rohit Prasad, engineer behind Amazon Alexa, stated that the Turing test is no longer sufficient to measure the success of artificial intelligence. Stating that artificial intelligence is much more advanced than a technology that tries to imitate humans, Prasad stated that a new test protocol is needed.
The Head of Alexa Artificial Intelligence Assistant, Rohit Prasad, stated that the Turing test is no longer sufficient to measure the impact of artificial intelligence, and a new evaluation protocol is needed.
Prasad stated that while artificial intelligence is far superior to technology that can only mimic a human, people are more satisfied with a technology that performs its duties rather than an artificial intelligence close to them.
Artificial intelligence was expected to become indistinguishable from humans
The purpose of the Turing test is to detect whether the person with whom a person is chatting is artificial intelligence. In the test process, an artificial intelligence and a real person are placed in front of the person making the speech. By asking certain questions, the speaker tries to determine which of the answers on the computer screen belongs to artificial intelligence and which belongs to a real person.
British mathematician and computer scientist Alan Turing predicted that by 2000, humans would have less than 70% chance of distinguishing artificial intelligence from humans thanks to the Turing test. However, the technology industry has yet to achieve such success in artificial intelligence.
People prefer AI to be functional
Rohit Prasad, Head of Alexa Artificial Intelligence Assistant Division at Amazon, wrote in a magazine article that the Turing test was full of limitations and that this test did not provide any benefit in the researches scientists are working on.
According to Prasad, people have already lost interest in differentiating artificial intelligence. Instead, they expect the AIs found in their homes, cars, and even smartphones to be functional and transparent.
Prasad states that Turing tests take artificial intelligence technology very lightly. Stating that waiting for artificial intelligence to give humane answers during the Turing test was a deception, Prasad made the following statements on the subject; When asked questions such as, “How far is zone A from zone B? To artificial intelligence? In order for it to appear humane, it has to answer with hesitation. Artificial intelligence tries to prove that it is close to human beings by pretending to be thinking and responding after a while. However, such tests, It does not take into account the sensors and skills that artificial intelligence uses to see, perceive and experience the outside world. “
“We should focus on artificial intelligence that will increase human intelligence and quality of life”
Stating that artificial intelligence does not need such tests to increase its functionality, Prasad said; “If you ask the AI to turn the lights off, you would rather just make short sentences rather than having a lengthy dialogue with it. Even when you ask artificial intelligence to tell your child stories, you want to be sure that the counterpart is artificial intelligence in a transparent way.
“Instead of being obsessed with distinguishing artificial intelligence from humans, we need to focus more on artificial intelligence that will improve human intelligence and quality of life. In line with this goal, common sense, language competence, performing tasks instead of people and making quick calculations are the first issues to be taken into account while developing artificial intelligence. “
Artificial intelligence will be evaluated with wider scope
Prasad stated that in 2016, with a competition called Alexa Prize Socialbot Grand Challenge, the best social assistant competition was organized, and thanks to this competition, artificial intelligence was evaluated not only with the Turing test, but with a much wider scope.
The aim of this competition was to find social assistants who could have fun and pleasant conversations rather than convince people that artificial intelligence is a human being. During the competition, each social bot was asked to chat with the jury for 20 minutes on many different topics such as sports, technology, and politics. The juries voted for the social bots they deemed worth chatting again, allowing the bots to go to the final round. Thanks to this competition, it has been revealed that artificial intelligence can develop their sense of humor without the need to pretend to be human where necessary.
In addition to performing simple tasks such as setting an alarm, AI Prasad says they should also have analysis capabilities and be able to identify our needs ahead of time.
Prasad gave an example of this feature that artificial intelligence should have with the following words; “After setting the alarm, if you go to the bedroom without turning off the lights downstairs, Alexa can warn you that you have to turn the lights off. The same happens when Alexa detects a breaking sound. Alexa notices that a glass is broken inside the house, and advises you to take action by warning. “