How Automotive AI Is Being Used to Make New Cars Even Better
Artificial intelligence is doing its damndest to change everything, and what you drive is no exception.
AI is everywhere. Or it will be soon, at least, if you listen to the endless techy pundits predicting the future. Jobs have been lost, lawsuits filed, and there's a general sense and anxiety that everything is about to change in a big way, thanks to this looming trend that few are excited about and even fewer truly understand.
0:00 / 0:00
But AI isn't one specific thing. Artificial intelligence is a broad, umbrella term covering several technology-related topics that have been in motion for decades. And while AI chatbots and automated image generation are getting the most buzz, there are endless applications for this technology. An increasing number of those fall within the automotive space, including some you might not have expected.
But before we look at those, let's take a moment to define what artificial intelligence is.
What is Artificial Intelligence?
Artificial intelligence, or AI, is quite simply when some artificial system exhibits a degree of intelligence, meaning it can learn and make decisions based on that learning. Typically, these days, we're talking about a piece of software or computer hardware.
Now, that's again a very broad definition, but believe it or not, there isn't much in the way of formal guidelines for what is and is not intelligence. For a long time, the so-called Turing Test, or the imitation game, was believed to be the gold standard for determining whether an AI had reached true intelligence.
The test, proposed by English mathematician and overall genius Alan Turing in the 1950s, simply describes a scenario where a human being has a conversation with two other entities. One of those entities is a human, the other is an AI agent. If the person asking the questions can't tell which is human and which is artificial, then the AI is said to have passed the test.
For decades, that was such an impossible-seeming task that it quickly elevated to the gold standard. Now, however, numerous AI chatbots can fool many people, leaving us still in doubt about what AI is.
Beyond that, we're seeing more and more AI applications that are well beyond simply having conversations. Additionally, there are many different technologies at play here.
One of the most significant is called machine learning, or ML. This is simply the ability of a digital system to teach itself. Most automated driving systems rely on ML to some degree, watching millions of hours of recorded driving footage and picking up correct behavior. But while a system that uses ML is learning on its own and making decisions based on that learning, you wouldn't necessarily call one of those systems an AI.
Rise of the In-Car Chatbots
We'll start with the most familiar applications of AI: virtual agents you can speak with. Endless automotive manufacturers have partnered with companies like OpenAI to bring these systems into the car, including Mercedes-Benz, Volkswagen, and more.
The basic idea is that you can have a conversation, ask questions, get more nuanced information on where you are, and even get them to tell you a story if you're really bored on a long drive. But that's all the same stuff you can do with OpenAI anywhere, and frankly, it's not a massive leap beyond Siri, who we've been chatting with in our cars for a decade now.
But other in-car AI agents are coming with a bit more purpose. BMW's implementation, created in partnership with Amazon's Large Language Model, has a single goal: Make confusing cars easier to use.
First demonstrated at this year's CES in Las Vegas, the system is basically a custom AI agent that was only trained up on BMW's documentation surrounding your car. In other words, it knows everything about what your car can and can't do and can not only change car settings for you but also explain to you what those settings mean and even when you'd want to use them.
One example would be changing driving modes. Say you're driving a new BMW M4 CS and going out for a spirited drive, but you're driving on some rough roads, so you want the suspension to be soft. You could use the touchscreen to dig through and create a custom drive mode just for the day. But with this system, you could say, “I’m going to drive on some twisty, bumpy roads. Can you make me a custom drive mode?”
The car will ask you a few questions about what you want and then apply the changes, all without you having to take your hands off the wheel.
Better Range Through AI
These AI systems aren't just things that we'll interact with. An increasing number of artificially intelligent systems do their work on your car long before it even shows up on the dealership lot. Some of the more interesting applications of this technology are scoring us more range in our EVs.
One is with a company called Chemix, which we profiled earlier in our piece on EV battery life. While EV batteries are well-understood things from a technology standpoint, there is still a lot of nuance to the blend of elements that compose your average battery pack.
Endless permutations of different cathode, anode, and electrolyte construction can have a significant impact on a battery’s ability to perform in cold temperatures and hot temperatures, even impacting its recharge speed.
Chemix's system can virtually cycle through those permutations to identify potential combinations that address whatever goals the manufacturer has in mind for their next-gen battery pack—all without having to build real batteries for every test.
Chemix has said its technology has already derived new chemistries that can double the number of charge cycles a given battery can handle, which might make those eight-year battery warranties seem a little unnecessary.
And similar efforts are afoot to optimize other internal design systems for EVs. A company called Neural Concept is applying ML to the science of aerodynamic design. One of the most complex and important aspects of aerodynamic design entails computer flow simulations. Basically, digital wind tunnels.
These traditional systems require huge computers running for hours at a time to figure out the math involved, but Neural Concept has found a way to do it using ML. The company's technology, called Neural Concept Shape, can very quickly estimate what would be an improved design for a given part, even battery cooling plates, which would make batteries more efficient.
Parts like that will help give your next EV more range, but there are other applications, too. Four current Formula 1 teams rely on Neural Concept to help them improve the aerodynamic efficiency of their cars without using any of their precious wind tunnel time.
Off-Road Self-Driving
By now, you've surely heard plenty about self-driving cars and how they'll soon whisk you to work and back while you catch up on a little sleep, technology that's perpetually just a few years away. Self-driving cars are generally trained using ML exhibiting some elements of AI, but the basic concept of self-driving is hardly new or novel.
What is new, however, is taking some of those same concepts and applying them to off-road motoring. Yes, self-driving all-terrain vehicle technology is being developed by a Canadian company called Potential Motors.
Potential is developing what it calls Terrain Intelligence, which relies on a fleet of sensors to view the world around it. Unlike on-road autonomy, there aren't any lane markings to follow or rules of the road to obey. Things also tend to be a little less predictable.
I know what you're thinking: What's the point of off-roading if you're not driving? The idea isn't necessarily to hand over control but to rely on the system to get better guidance about the terrain around you. Think of it like a virtual spotter who won't complain about having to go wade through the muck.
Potential is working on deploying the technology with partner CF Moto on the company's side-by-sides but is also developing its own little purpose-built off-roader called Adventure 1.
Sound Check
Getting truly great sound in a car depends on many more factors than big speakers connected to a quality amplifier. The sheer shape of the car's interior has a massive effect on sound quality, on top of things like ambient noise intrusion and even the number and preferences of the people inside the car.
With EVs gaining market share, in-car audio is only getting more important. Ditching the noisy combustion cycle from the mix creates a soundscape that's far more conducive to great sound quality, which just adds more and more pressure on manufacturers to get it right.
As part of its Automotive Sound System, Yamaha has developed something called Music:AI that can optimize in-car audio, not only creating default settings that are better-calibrated for a given car but also dynamically tuning them to match the proclivities of its occupants.
Yamaha said this was trained based on the knowledge of its sound experts, allowing the AI to quickly and dynamically tune in a given sound system for a given interior. It can then modify that audio configuration dynamically, depending on the nature of the music you're listening to.
Beyond that, it can converse with the people in the car to get their feedback on how it sounds to them, adjusting on the fly. It sounds like it could be a subtle but amazing improvement for in-car audio, and Yamaha promises it'll be available in cars as soon as next year.
Naturally, this will be integrated into Yamaha's amplifiers, so other manufacturers will be out of luck. But if it works well enough, hopefully others will come up with their own implementations of similar tech.
High-Quality Inspections
It can be a little nerve-wracking to drop your car off at a service station to get some work done. You never know exactly how well it's going to be treated, and we've all seen or heard horror stories about cars coming back with curbed wheels or scuffed fenders that weren't there before.
UVEye's automated inspection system is designed to prevent any debate on issues like that. It's basically a high-resolution set of cameras set up in a drive-through portal that looks like something capable of teleporting you to another dimension.
It doesn't do that. Instead, it scans every exterior surface of your car, creating an immediate record of how things looked when you rolled in. That's nice for liability purposes, but that's not what makes this interesting.
The intelligent part happens on the back end. That full scan of your car is fed into a learning system that has been taught to automatically identify trouble spots. It can spot bald tires, for example, broken exhaust parts, and even things like oil leaks, all automatically.
It takes that information and generates a detailed report that, interestingly, doesn't just go to the service station to upsell you on whatever recall you really bought the car in for. The owner of the car also gets a full copy of the report, the idea being that you can take that information and cross-shop for any repair work that needs doing.
But mostly, you can have a little proof that the lip on your polished wheels was immaculate when you rolled in.
More to Come
And all this is just the beginning. Although many have been in development for years, AI technologies are progressing at an incredible rate. Meanwhile, the companies developing them are seeing investment of the sorts not seen since the early days of the cryptocurrency boom.
Exactly what shape all this will take and whether any of these technologies can really deliver on their potential remains to be seen. All we know is that if we get a compelling version of KITT in our next car, we'll be very happy indeed.
