Will Artificial Intelligence Become Your Co-Pilot on the Road? (.PDF Download)
Tractica opened eyes with its detailed market forecasts for artificial-intelligence (AI) hardware, software, and services targeting the automotive market during the 2016 through 2025 period. The report defines AI as a technology that uses data and algorithms to mimic an individual’s ability to learn and solve problems. Tractica says the automotive industry has seen the promise of such technology, and is among the industries at the forefront of using AI to augment human actions and mimic the actions of humans, while also harnessing the advanced reaction times and pinpoint precision of machine-based systems.
Both semi-autonomous and fully autonomous vehicles of the future will rely heavily on AI systems. However, associated AI algorithms can require enormous resources of memory and computer time. Therefore, efficient AI algorithms are a high priority.
Most AI hardware and software includes a mixture of various technologies. One that’s particularly critical is deep learning, which imitates the activity of the brain’s ability to think. This software learns, in a very real sense, to recognize patterns in digital representations of sounds, images, and other data.
In fact, deep-learning technology is expected to be one of the biggest and the fastest-growing areas in automotive AI. It’s used in voice recognition, voice search, recommendation engines, sentiment analysis, image recognition, and motion detection. Autonomous cars employ deep-learning technology for image processing, speech recognition, and data analysis.
It’s difficult to put a number on the amount of code that has to be written and the amount of memory that will be required for AI. There’s no doubt, though, it will require mega man-hours to write the necessary code.
Thousands of lines of code present a dilemma to automotive electronics system designers. They would undoubtedly have to support these systems with periodic updates for the lifecycle of the car. And, large code bases are prone to introducing errors. There may be a compromise to minimize dependence on software, which would require special hardware circuits or systems on a chip (SoCs) to take over some of the software tasks. However, this could exacerbate infighting between hardware and software engineers trying to solve the cause of problems.
Another subject that must be resolved is what processor and what speed will be required to handle AI in a timely manner. Then, how much power will be needed for the processor and memory? In an EV, this could siphon power away from the battery that supplies power to the vehicle. And, how will the necessary circuits be packaged so that they can fit in the car and allow them to be maintained?