Home SECURITY People are no longer needed – robots have learned to train themselves

People are no longer needed – robots have learned to train themselves

0
People are no longer needed – robots have learned to train themselves

[ad_1]

People are no longer needed – robots have learned to train themselves

The teacher-student model reduces AI training time by 100 times.

University of Southern California study shows that AI models that share knowledge among themselves can learn from each other and quickly master various tasks. The discovery has the potential for application in various fields, including medicine.

A team of researchers have developed a tool called SKILL (Shared Knowledge Lifelong Learning), which helped AI learn 102 different tasks – for example, categorizing tens of thousands of car images by model, colors by species, chest x-rays by diseases. Then different AI models shared their knowledge through a decentralized communication network, and eventually all models mastered the knowledge of all 102 tasks.


In simple words: you are a poker master, your friend is a connoisseur of French cuisine, and the third friend is an expert in Mozart’s music. You meet and share knowledge in your fields. As a result, each of them learns something new from the other two.

SKILL includes a set of algorithms that speed up the learning process because AI agents learn at the same time in parallel. The study showed that if each of the 102 AI agents learned one task and then shared it with other agents, the amount of training time needed would decrease by a factor of 101.5, taking into account the necessary communication and knowledge consolidation between agents.

The study can be represented as follows: each AI teaches a lesson in its specialty, and all other robots are attentive students. They share knowledge through a digital network that connects them all, like their own private internet.

Scientists believe that SKILL is a promising starting point for progress in the field of Lifelong Learning (LL). According to experts, this research could be scaled up to thousands or millions of tasks in the future.

In the medical field, various AI systems can specialize in the study of various diseases, treatments, patient care techniques, and the latest research. After consolidating their knowledge, AI can serve as a comprehensive medical assistant, providing doctors with up-to-date and accurate information in all areas of medicine.

[ad_2]

Source link

www.securitylab.ru

LEAVE A REPLY

Please enter your comment!
Please enter your name here