麻豆淫院

A team of computer scientists, AI specialists and veterinary medicine researchers in the U.S. and the U.K., working with the dog training center, The Seeing Eye, Inc., has found that an AI model was better than humans at assessing puppies for characteristics of good service dogs.

In their study, in the journal Applied Animal Behaviour Science, the group asked expert dog trainers to track attributes of puppies and adult dogs as they progressed through a service dog training regimen and used that data to train an AI model.

Prior research has shown that there are several traits inherent in some puppies that make them well suited to training as service dogs. But because there are so many traits, and because of the uniqueness of individual animals, choosing the right puppy is oftentimes more art than science. In this new effort, the research team attempted to apply science to the problem to reduce the number of failures.

Trainers at Seeing Eye kept logs describing characteristics or traits of the dogs as they took them through the training process, noting most specifically which characteristics seemed to lead to a successful outcome: an adult dog with all the traits required to perform successfully as a service dog. Each of the trainers also filled out periodic questionnaires regarding the dogs' personalities, temperament and focus.

The researchers then used that data to train an AI model to be used for puppy assessment. They used the model to make predictions of puppies regarding their suitability to serve as a seeing-eye-dog.

After a year of testing, the researchers compared the results of the AI models to those of humans who had been trained to pick out puppies and found that the AI model was more accurate鈥攐ne model even achieved a success rate of 80%.

漏 2025 Science X Network