ChatGPT in Cars: The End of Touch
Multiple auto makers have brought ChatGPT to their dashboards, but announcements at CES in Las Vegas a week ago left observers wondering whether auto makers are asking the right questions.? With few exceptions, auto-centric ChatGPT announcements appeared to focus on accessing the owner’s manual, which is a far cry from a high frequency application.
Volkswagen’s press event was perhaps the most amusing.? The presenter suggested asking ChatGPT to tell a story about dynosaurs.? This seemed to be missing the point.
The point of ChatGPT-style generative AI solutions is to enable access to relevant data sets linked to a familiar voice recognition front end to convey urgently needed information to drivers – particularly regarding driving circumstances.? Even more important, generative AI solutions have the potential to reduce or eliminate the distraction of screen touching – marking a critical industry turning point.
(This isn’t to suggest that touch screens will go away.? It is merely to suggest that screen touching will be less necessary.)
General Motors was arguably the first car maker to announce its intention to integrate ChatGPT.? Mercedes-Benz was the first to make a beta version available to a limited set of vehicle owners.? Stellantis was the first to market, delivering a ChatGPT-enabled voice recognition interface from SoundHound to millions of cars already on the road in Europe via an over-the-air software update.? Volkswagen made the first formal announcement of its ChatGPT plans at CES with Cerence.
None of these announcements has captured the use case scenario that I have been envisioning for ChatGPT.? The circumstance that tends to make me more crazy than usual is when my BMW embedded navigation offers me a detour from my existing navigation route and asks me to accept the route change.? For me, this is a total “WTF” moment.? How am I supposed to make this decision while I am typically driving 60mph?
With ChatGPT on board, I could ask: “Can you explain my routing options?” or “Why are you recommending a detour?”? Pretty simple stuff – but very useful.
With nasty winter weather returning to my Northern Virginia neighborhood after a two-year hiatus I had lots of questions for a potential ChatGPT driving partner in recent days:? “How safe are the roads?”? “What is the safest route?”
Traffic data expert Inrix has taken on this challenge offering a generative AI solution for municipal planners and traffic executives.? Called “Compass,” the Inrix solution is designed to tap into the company’s “50 petabytes” of transportation data collected over the past 20 years – from a wide range of sources including connected vehicles, trucks, mobile devices, incident and weather reports, traffic cameras, smart infrastructure, events, and more, the company says.
The questions that Inrix is focusing on include:
“What is the best way to reduce traffic on this road during rush hour?”
领英推荐
“Which are the most dangerous intersections?”
“Why is this intersection so dangerous?”
“How can I make this intersection safer?”
“What will happen to congestion if I add a bike lane?”
Inrix sees Compass as a tool for identifying problems, finding solutions, and discovering causes or anticipating alternatives.? These are precisely the challenges facing drivers.
ChatGPT could become the ultimate wingman (or wing-entity) for overcoming anxiety while driving.? Why is this road backed up?? Are there alternative routes?? What is the best time to drive from point A to point B – i.e. when will the drive time be the shortest?
The best embedded navigation and infotainment systems are designed to reduce driver stress.? If ChatGPT-style generative AI solutions – with familiar voice recognition front ends – can better explain the driving context, drivers will be better able to relax.
Most important of all, a voice-based solution will obviate the need to pay attention to the infotainment screen for traffic messages or on-map icons or construction or car crashes.? I know BMW is trying to be helpful with color coding, 3-D building animations, construction icons, and the like, but that’s just adding to the distraction.?
Ironically, BMW was one of the last car makers to adopt touch screens – preferring that drivers use the i-drive hardware controller.? BMW is now one of the worst offenders when it comes to encouraging screen touching.? The obvious corollary to touching the screen is taking your eyes off the road.
To be successful, ChatGPT-style generative AI solutions need to be infused with the right data – such as Inrix’s petabytes – enabled by natural language understanding (NLU) voice interfaces – and tuned to driving applications.? That’s a solution capable of saving time, money, lives, and aggravation. ?And it also means a lot less screen touching.
Application Support Engineer | L2 | L3 | Oracle | PL/SQL | Java | SQL | REST APIs | Integrations | Performance Tuning | Incident Management | Core Banking | Payments | Production Services | Fintech | Matera | Splunk
11 个月I believe that Artificial Intelligence can also help a lot with predictability factors for the driver. Situations that would perhaps be checked using cameras or sensors can be improved by using AI to your advantage. With the purpose of increasing safety, comfort and reducing interaction on screens.
Consulting Services
1 年Yes ChatGPT can be the LLM that does a nice job of conveying information, but there needs to be some more AI behind that to come up with the actual DATA. On your example of your navigation route change, the current data available and info communicated are woefully inadequate. Last summer on a fun Colorado road trip I got a Google Maps message that the road ahead was blocked and it offered a logical detour. Well, I was only 20 miles or so from my destination, and the detour added 2 hours. Um, did that "blocked" road clear sooner than that? Recently we had a huge avalanche on a nearby mountain pass that covered 10 cars (everyone got out ok). The detour to Denver was a 6 hour adventure and some people found the downstream Interstate leg on that route closed anyway. So the data needed is about the reason for the detour as you state, but also estimates of delay based on historical events, predictive traffic and fusing of diverse sources of info. Throw in V2X for good measure. And while you're at it chatbot, please tell me about that alternative route. I am excited that all of this will happen; we're just a long ways from that dream so far.
Bridging technology and business, sharing industry experience, caring about people and the planet.
1 年Roger in your article you hit the nail on the head! ChatGPT provides a great interface to ask questions relevant to my journey but it's not being used for that. Instead it's restricted to demos about dinosaurs by outdated APIs which were designed for graphical interfaces. I would love to be able to say "Plan a route to Edinburgh with a place to stop for dinner" with the expectation the car will plan the optimal route which has a stop around 18:00 where I can charge the car and refuel myself at my favourite type of restaurant. Soundhound, Cerence and others can easily implement the language understanding. What is missing today is the combination of the vehicle data, location data and my personal preferences to make that a really useful "wow" experience. Time we step up as an industry to make this happen. Then we will see the real value of voice assistants in cars.
CPO/CTO/COO, Board member, Investor
1 年Industries always take new (cool) tech and apply it to every (often well solved) problem making the solution considerably worse. Nothing beats a well positioned volume knob that can be operated with 100% predictably in <<1sec, a manual hand/parking brake in an emergency situation or a big red glowing hazard warning button center dashboard. Making super simple interactions complicated is often the outcome. Making complex interactions (navigate me to X while passing by a Y gas station that accepts Z payment) easy/simple is the true potential for voice and touch. until then it is still loads of fun to have a message being read out while settings are in a different language ...
AI Engineer | AI Agents & Automation | AI for SMEs | 7+ in AI/ML | VoiceAI | LLMs | RAG | Vector DBs | LangGraph | 17+ Full-stack Dev. | Python, JavaScript, TypeScript | Next.js, Node.js | Docker, Linux, GCP, AWS, Azure
1 年Voice interfaces will be the end of infotainment screens! ????