I would echo the concerns about the unrestricted use of AI.
While I would encourage teams to incorporate AI into their design, unrestricted use of AI-based sensors like the Husky Lens do not lead to meaningful learning. With devices like the OpenMV cam and libraries like OpenCV, the barrier to entry for teams creating their own machine vision solutions is already low. I can’t see how it can be beneficial to further lower the bar to the level of a “one-click” solution like the Husky Lens.
The rules says that “teams must demonstrate a decent understanding of the AI technologies they use”, but I have met many who can present themselves as knowledgeable about a topic (…whether it’s AI, machine vision, or math), and their lack of understanding is only apparent when they are tasked to implement it. If they truly understand the AI technology they are using, then there would be no need for them to rely on a manufactured solution like the Husky Lens. Besides, how could one even understand the AI technology being used by a closed source device like the Husky Lens? At best, we could do no more than make some guesses about their underlying algorithm.
Furthermore, what would be the treatment for teams who failed to demonstrate understanding of the AI technology they are using? Would they be disqualified? How would a judge assess this fairly? This seems no different from buying a sensor that is “specifically designed or marketed to complete any single primary task of RoboCupJunior Rescue”.