THE’Artificial intelligence returns to kindergarten: Like a child he needs to learn the letters before he can read or numbers before he can count. These systems need to start from the basesin this way they manage to learn much faster when you then move on to more complex tasks. This is suggested by the study of New York University published in the magazine Nature Machine Intelligence, which opens up to methods to improve the training of the IAs.
“The IAs must first face the kindergarten in order to learn better complex tasks,” says Cristina Savin, who coordinated the study together with Christine Constantinople. “Overall, these results indicate ways to improve learning in artificial intelligence systems – adds Savin – and require the development of a more holistic understanding of how past experiences influence the learning of new skills”.
The researchers started from laboratory experiments on rats: The animals were trained to look for water, knowing that its presence was associated with certain sounds and the lighting of lights on the openings of the labyrinth. To reach the reward, therefore, they had to first learn basic information and then combine it together to complete a complex activity. The results they were then transferred to the training of the so -called ‘Recurring neural networks’, AI -based systems and designed to process information based on the knowledge already stored, particularly useful in vocal recognition and translations. Compared with those trained with normally used methods, the networks that followed the ‘asylum method’ learned much faster, thus becoming more efficient.
Reserved reproduction © Copyright Ansa
Source: Ansa
I am a journalist with over 6 years of experience working in the news industry. I currently work as an author for Global Happenings, and my coverage focuses on Technology news. I have written for various publications, including Reuters, The New York Times and The Guardian.
