Whether you call it Industry 4.0 or the Fourth Industrial Revolution, the next phase of manufacturing is here. This new era is pushing the future of AI and robotics to the forefront along with shop floor technology. While these innovations are important to help usher in enhanced productivity and flexibility to manufacturing, I have a few predictions on how how the future of AI and robotics will play out:
Embodied AI
There are various manufacturing tasks with their intricacies. Even the same task can differ when performed on a different shop floor. That makes rare edge cases inevitable. When rare edge cases stack up, that’s when the overall performance of the system degrades. To address this, we must move on from heuristics and hand-crafted models with engineered optimization objectives. The ability to deploy a fleet of robots, having access to vast amounts of data, and ‘learning on the job’ without supervision, will enable the robots to continuously learn on their own and tackle those rare edges. The key is to allow AI to physically interact with the environment and learn on its own. Because of that, I think embodied AI will become more prevalent.
Reasoning in AI
I believe AI will make meaningful progress specifically in reasoning. With a breakthrough in reasoning and planning for long-horizon tasks, robots will be able to perform tasks better than humans. The goal is not to be as good as humans or imitate what humans do, the goal is to come up with better ways of doing a task. That’s when emergent behaviors start to appear.
Bring Your Own Data (BYOD)
Today, robots are programmed through code and algorithms. We see code/algorithm and data as two different entities but I envision this clear boundary between data and code will start to disappear. Instead of using a programming language to describe a task, we will just show or describe what our intention is and our robots will learn that without any explicit model or algorithm development.
Overall, I see Industry 4.0 making robotics more affordable and cost-effective. Think about the first calculator; in 1972 it launched with a price tag of $395 (which in today’s dollars would be roughly $2,948). As technology evolved and its application became more widespread, the price became more accessible. Once the technology reaches the point of cost accessibility it doesn’t make it an investment-based decision, but rather a decision that simply makes sense.
Nima Gard a scientist and engineer focused on developing AI for robotics. As the Director of Artificial Intelligence at Path Robotics, he’s serious about leveraging AI to support manufacturing.
