LG Electronics on Monday said that at least one independent study verified that its Roboking Turbo Plus vacuum cleaner is as intelligent as a six or seven-year-old child when it comes to deciding how to handle obstacles, where to go, and how to clean what needs cleaning. LG cited research done by Seoul National University’s Robotics & Intelligent Systems Lab to demonstrate the intelligence of the AI-powered vacuum cleaner. The study consisted of measurements of 100 different metrics and tested multiple robot vacuum cleaners that boast autonomous capabilities. The vacuums could fall into one of three categories, being labeled as a dolphin, ape, or a child. The university researchers confirmed that LG’s Roboking Turbo Plus vacuum features task-oriented intelligence comparable to that of a grade-schooler.
LG’s homegrown DeepThinQTM AI program powers the vacuum in question. A previous vacuum, released in 2015, managed to achieve ape-level intelligence in similar testing, and the Roboking Turbo Plus gained a leg up on its predecessor thanks to the deep learning algorithms present in LG’s DeepThinQTM program. The vacuum comes from the factory with a good amount of knowledge about how to handle certain types of environmental hazards but learns more over time simply by bumping into things or recording other adverse reactions to the ways it handles some obstacles. The vacuum is capable of figuring out what kind of an obstacle is in front of it, such as a piece of furniture, sleeping animal, or child, then deciding whether to simply power through, go around the obstacle, or stop and give the obstacle time to remove itself from the vacuum’s path.
LG’s DeepThinQTM is an example of the efficiency inherent in using AI for what it’s inherently best at – a singular task made up of many small tasks. All the vacuum wants to do is get around, but to do so, it has to see where it is in a room, see what’s in front of it, identify obstacles, and remember approaches to obstacles that didn’t work out in the past. At its core, this approach can be applied to almost any field of AI research, though with additional complications depending on the type of AI being used. General artificial intelligence, for example, takes that approach and ennobles it with some preloaded knowledge and training, as well as additional sensory capabilities. As the AI encounters bad reactions, it learns to identify good ones and what causes them. In some cases, AI programs built on this principle are also able to closely observe their users in order to facilitate faster and more in-depth learning.