Colorado State University researchers recently published a paper on an artificial intelligence (AI) system to train dogs. The researchers used AI to train the software on thousands of canine images. The hope is that computers could one day be used for an automated dog training system. “One of the ways that AI excels is in the ability to perform repetitive tasks uninterruptedly,” Cady Bocum, co-founder and CEO of AI solutions provider GBSN Research, who was not involved in the research, said in an email interview. “When it comes to dog training, patience, and perseverance go hand in hand.”
Sit, Stand, Lie Down
The paper’s authors, computer science students Jason Stock and Tom Cavey, used software to determine whether a dog is sitting, standing, or lying. If a dog responds to a command by adopting the correct posture, the machine dispenses a treat. “Dog trainers have mastered the skill of teaching obedience by rewarding desired actions with food or auditory queues,” Stock and Cavey wrote in their paper. However, the authors continued, “The learned behavior of dogs may diminish in early stages of training when actions go unrewarded. To maximize the efficacy of learning, we model the actions of a dog trainer with machine learning to identify behaviors and reinforce commands such as ‘sit’ or ’lie down’ in real-time.” The researchers tested their models on Henry, Cavey’s Australian Shepherd, according to a news release.
Fetching Dog Data
To train the software, the researchers needed to find images showing dogs in various postures. They found what they were looking for in the Stanford Dogs dataset, an Internet collection that contains images of 120 breeds of dogs from around the world. All told, there were more than 20,000 images, showing many positions and various sizes, though it required preprocessing. The researchers wrote a program to help quickly label the images. But why use AI for dog training at all? Because it relies more on evidence-based data than theory, John Suit, advising chief technology officer at robot dog firm KODA, said in an email interview. “One of the main qualities humans have over AI is our capacity for empathy—that could work against a dog during its training,” he said. “By only focusing on rewarding positive behavior, AI systems that are learning how a dog behaves can train the dog in faster and more effective ways than a human could.” AI could be used for all kinds of canine behavior issues, speculated animal behavior expert Russell Hartstein in an email interview. “A dog that has just had surgery and is wearing a collar to stop him from gnawing at the stitches or wound, for example,” he said. “AI would be helpful to alert a parent if/when a dog is doing a problematic behavior.”
AI Training for Humans, Too
It’s not just dogs that can benefit from AI training, Javed Ahmed, the senior data scientist for analytics education company Metis, said in an email interview. “As the tools get more powerful, the breadth of application is expanding beyond human factors,” Ahmed said. “Tools from AI are almost ready to generate content that can be directly incorporated into training programs. For example, we may be able to auto-generate explanations for technical topics or multiple-choice questions for assessments.” Suit said AI could be particularly useful in training employees, too. “During the onboarding process, AI (through chatbots) can lead new hires through courses and employee handbooks,” he said. “This leads to more personalized learning and educating, while also freeing time for hiring managers and HR.” AI is being used in classrooms, as well, for tutoring, creating personalized student plans, grading, and managing homework, Bocum said. Digital learning platforms are increasingly popular in China, where students use them to boost their academic performance. “A high number of users is also one of the keys to the success of this model in the country: AI also needs practice,” she added. “The type of practice that comes from the data gathered by the millions of students using the system.”
