What’s Next in Generative AI: Key Takeaways on ChatGPT for Business from OpenAI’s COO
1:25 p.m. PT – One of the largest sessions at Nvidia GTC 2024 was a one-on-one conversation between OpenAI chief operating officer Brad Lightcap and Nvidia vice president of enterprise computing Manuvir Das. The focus of GTC and the future of Nvidia is primarily centered around the evolving landscape of generative artificial intelligence (gen AI), spearheaded by OpenAI and its groundbreaking ChatGPT chatbot. How groundbreaking, you ask? According to Lightcap, “90 percent of Fortune 500 companies are using ChatGPT in some form.” If that caught your attention, here are four additional insights from Lightcap and Das:
- It’s very early. This is especially pertinent for gen AI as a business (enterprise) tool. ChatGPT captivated audiences with its launch in November of 2022. According to Lightcap, “We are still on the flat part of the curve—really early in the phase shift. Companies are still rigging and pipelining the infrastructure.”
- On target setting: Lightcap advises, “Don’t overshoot; don’t try to swallow the ocean. But also don’t undershoot either. Don’t lack ambition.” Companies should begin by clearly defining the problem at hand. Only then should AI be applied to evaluate the results and their value.
- Democratize access: He recommends that leaders provide wide access to AI tools, as employees often understand the problems best and can organically discover solutions quickly, prioritizing this approach over a laborious process of perfecting a chatbot or other AI tool.
- Start with the gnarliest problem: Lightcap highlighted that customer service is an excellent domain for applying AI, given that it is a challenge across various industries. “No [company] likes the quality [of their customer service]. They spend a lot of money on it. And it doesn’t fix the problem,” he emphasized.
While specific car brands and the automotive sector were not directly referenced during this discussion, it is noteworthy that numerous automobile manufacturers feature on Fortune’s list of the top 500 global companies (such as Toyota, Volkswagen, Stellantis, etc.), making it easy to visualize the application of Lightcap’s insights on gen AI. –Ed Loh
Would You Trust a Self-Driving Car More If It Told You What It Was Thinking?
12:55 pm PT – The England-based autonomy startup Wayve is developing an AI driving system that learns the rules of the road through real-world experience instead of relying solely on fixed instructions. This innovative approach, however, is complemented by another novel concept aimed at enhancing passenger comfort during self-driving experiences.
Wayve’s LINGO-1 functions as a driving commentator, explaining the autonomous system’s actions, such as slowing down or turning. It can inform passengers with statements like, “I’m stopping for the pedestrian at the crosswalk” or “I’m slowing down for the upcoming left turn.” This program integrates images and driving data with rationale drawn from a large-language model in what Wayve dubs a visual-language-action model (VLAM).
Additionally, Wayve has created a text-based query system allowing passengers to inquire further into the AI driver’s decisions and observations. For instance, passengers can request Identifying the top three hazards the AI is tracking or clarification on why the vehicle has halted despite an unobstructed road ahead. Such transparency may alleviate broad public skepticism towards autonomous driving technology. —Eric Tingwall
3 Ways the 2024 Volvo EX90 Previews the Future of the Car
10:05 a.m. PT – I recently conversed with Alwin Bakkenes, Volvo’s head of software, regarding how the upcoming EX90 electric SUV distinguishes itself from Volvos currently on the market. His insights underscore the unique attributes that make a software-defined vehicle fundamentally different from traditional vehicles that might utilize software without being designed with it in mind.
- Central compute architecture: Most vehicles today are built on an electronic architecture grounded in domain controllers—where crucial functions (also known as domains), such as powertrain and chassis systems, operate on individual computers. The EX90 adopts a central compute or zonal architecture, harnessing Nvidia Orin and Xavier computers to facilitate comprehensive system management, save for one exception (details to follow). This centralized framework minimizes wiring, increases inter-system connectivity, and streamlines the updating process for any vehicle aspect in the future.
- Infotainment as a product: The EX90’s infotainment system is powered by a dedicated Qualcomm Snapdragon computer, independently operating from the Nvidia central compute architecture. This design choice stems from Volvo’s creation of the latest infotainment system as a standalone product, enabling its use across various models regardless of the vehicle’s underlying computing architecture. Consequently, this strategy simplifies the workload for software engineers, consolidating into a singular hardware-software “super set” utilized across all vehicles going forward, potentially lessening security risks and maintenance for infotainment updates.
You’re part of the development team: Scheduled for mid-2024, the EX90 will be equipped with the essential hardware to facilitate Volvo’s Ride Pilot, an eyes-off-the-road highway driving assistant. However, this feature will not be available upon launch. Volvo plans to roll it out as a subscription service for California drivers later on. Engineers require additional test data to proceed, which will be gathered as early EX90 owners operate their cars. The lidar, radar, and cameras will collect valuable data to validate Ride Pilot prior to its official release. –Eric Tingwall
Generative AI Is Transforming Car Design
9:44 a.m. PT – During a demonstration, we explored how generative AI is revolutionizing vehicle design. A notable insight was that this new wave of design tools can learn a designer’s individual style, enabling seemingly straightforward line drawings to be transformed into refined outputs. The AI can even be configured for varying brands or design languages, permitting a simple sketch (demonstrated as a basic doodle) to be expressed in the specific style of designer X, should the AI have been trained with selected images beforehand. Conversely, with different training images, the same design could emerge resembling a Porsche or even a Lamborghini.
Designers will gain the ability to quickly iterate on specific vehicle components. For instance, we were shown the original Ford Bronco model, humorously referred to as a “Jeep” by the AI team, and individual parts (such as grilles, wheels, and headlights) could effortlessly be exchanged with just a few clicks. Notably, vehicles such as the current Bronco, Ford F-150, or Toyota Land Cruiser maintain consistent body panels while their grilles change based on model variants, allowing designers to realize the profound capabilities AI offers in guiding refinements, facilitating rapid prototyping.
Additionally, AI also enhances vehicle rendering capabilities. While generating an image of a car is critical, rendering it in a realistic 3D environment is of equal importance. We observed a scenario where a designer crafted a darkly atmospheric garage, initially in 2D format. The AI program transformed this basic design into an accurately illuminated 3D space, featuring a parked vehicle. This compelling technology not only aids current designers but also serves as a powerful tool for inspiring upcoming generations passionate about automotive design. –Jonny Lieberman
What Car Enthusiasts Need to Know About Nvidia GTC 2024
9:24 a.m. PT – For car aficionados intrigued by the intricate technologies beneath the hood, understanding computing fundamentals has become imperative. No longer can you delve into automotive technology without recognizing software’s vital role. Here is a primer of essential terminology you’ll encounter frequently in our coverage of Nvidia’s GTC 2024.
- What’s a software-defined vehicle? Electric vehicles (EVs) and autonomous driving systems are garnering attention, yet both hinge on the ongoing transformation towards software-defined vehicles (SDVs). An SDV is characterized by interactions and key functionality significantly influenced by software. For instance, with an SDV, drivers might unlock their cars using their smartphones, temporarily upgrade horsepower for a day at the track, or summon remote diagnostics for peculiar vibrations. The Tesla Model S, launched in 2012, is widely recognized as the pioneering SDV, and Tesla continues to set the benchmark for automakers advancing their SDV initiatives.
- What is Nvidia Drive Thor? Drive Thor is a state-of-the-art central computing unit by Nvidia designed to manage autonomous driving, driver-assistance functions, and infotainment—all employed by various automakers and self-driving tech developers. This System-on-a-Chip (SoC) amalgamates CPUs and GPUs, boasting a performance rating of 1,000 teraflops (TFLOPS), delivering fourfold the capabilities of its predecessor, Drive Orin. For context, a teraflop equals a trillion floating-point operations, meaning a computer with 1,000 teraflops achieves about 80 times the performance of a Microsoft Xbox Series X.
- What’s a CPU? A central processing unit functions as a computer’s brain, managing instructions and coordinating tasks across all components to ensure operations occur sequentially, with minimal delay, maximizing efficiency.
- What’s a GPU? Originally aimed at improving visual quality, graphics processing units are now integral for executing complex tasks such as training and applying AI algorithms. While CPUs excel at executing one instruction at a time, GPUs are adept at processing multiple commands concurrently.
- What is Nvidia Omniverse? This 3D graphics software is utilized by automotive manufacturers for diverse applications, from running online vehicle configurators to simulating entire assembly line operations. Omniverse can be conceptualized as Nvidia’s platform for crafting the metaverse. Rather than creating an escape filled with fragmented avatars, auto makers harness it to create digital twins that depict their vehicles and elucidate real-world functionalities.
- What’s a digital twin? A digital twin represents a virtual model of a physically existing entity. It often emphasizes precise details crucial for emulating complex, dynamic environments—such as manufacturing plants. Our detailed exploration of digital twins discusses how enterprises like BMW are employing this advanced technology today. –Eric Tingwall
How AI Supercomputing Power Is Coming to Your Car
Nvidia Founder/CEO Jensen Huang Announces Blackwell GPU and New Drive Thor Partnerships with BYD.
9:00 a.m. PT – For 15 years, Nvidia Corporation has organized GTC (the GPU Technology Conference), but it has been five years since this conference was conducted live. Graphics processing units (GPUs) are pivotal in rendering realistic lighting and visual effects in video games, forming the foundation of Nvidia’s success over the past 30 years. Established in a humble booth of a Denny’s in Silicon Valley back in 1993, Nvidia is currently valued at over two trillion dollars, placing it among a select few firms surpassing a trillion-dollar market capitalization.
At GTC 2024, Nvidia is shining a light on artificial intelligence (AI) across three fronts: automotive, robotics, and a 3D graphics platform termed “Omniverse,” which co-founder Jensen Huang asserted provided the stunning video simulations displayed during his lengthy keynote address.
Huang initiated his presentation with a brief recounting of Nvidia’s significant influence on AI as it exists today, addressing the thousands assembled at San Jose’s SAP arena. He recalled hand-delivering Nvidia’s latest GPU in 2016 to a startup called OpenAI. Fast forward to 2022, OpenAI unleashed ChatGPT, a launch that, according to Huang, “captured the world’s imagination” and “helped us realize the importance and capabilities of artificial intelligence.”
Utilizing this introduction, Huang unveiled Nvidia’s new powerful Blackwell GPU, recognized as the largest GPU crafted thus far. It is designed to be configured into an AI supercomputer (in conjunction with other Nvidia components like the Grace CPU, rack, and thermal management systems) to bolster the development of generative AI tools and systems by both Nvidia and its collaborators. Following Huang’s presentation, a press release from Nvidia listed several companies and executives endorsing Blackwell, including Alphabet and Google’s CEO Sundar Pichai, Amazon’s CEO Andy Jassy, Dell’s CEO Michael Dell, Meta’s CEO Mark Zuckerberg, Microsoft’s CEO Satya Nadella, OpenAI’s CEO Sam Altman, Oracle’s Chairman Larry Ellison, and Tesla/xAI’s CEO Elon Musk.
The only substantial automotive announcement occurred towards the conclusion of Huang’s address and was brief: BYD, recognized as the world’s leading EV producer, has selected Nvidia’s Drive Thor in-vehicle computing platform for its forthcoming vehicle lineup. A post-event press statement detailed BYD’s plan to utilize Nvidia’s Drive Thor for generative AI applications, which include the future digital cockpit, Advanced Driver Assistance Systems (ADAS), and autonomous driving functionalities.
BYD also intends to implement Nvidia technology in a “car to cloud” fashion, incorporating Nvidia’s Isaac robotics and Omniverse platforms to create “digital twins” that span every phase of vehicle manufacturing, from designing factories to enhancing retail experience and post-sale support.
BYD was not the only Chinese OEM announcing collaboration with Nvidia at GTC 2024. Xpeng shared plans for Drive Thor to power their XNGP AI-assisted driving system in their upcoming models, while Hyper, a luxury brand under GAC AION, revealed intentions to adopt Drive Thor beginning in 2025 to deliver Level 4 autonomous driving capabilities. These three Chinese brands join Li Auto and Zeekr, which previously declared their plans to integrate Drive Thor in future vehicles.
So, what implications does this hold for American consumers? At present, not much. While Nvidia’s supercomputing power is on the verge of becoming integral to vehicles—supporting all facets of planning, design, manufacturing, sales, and service—the latest AI-driven hardware and software will likely first debut in the streets of Beijing and Shanghai before making their way to Detroit or L.A. –Ed Loh
Edward Loh
I used to go kick tires with my dad at local car dealerships. I was the kid quizzing the sales guys on horsepower and 0-60 times while Dad wandered around undisturbed. When the salesmen finally cornered him, I’d grab as much of the glossy product literature as I could carry. One booklet that still stands out is the glossy overview of the Mitsubishi Eclipse GSX that favorably compared it to Porsches of that era. I would pore over each page, from the prose to pictures and specifications, never imagining I would one day be accountable for those asterisked figures “*as tested by Motor Trend magazine.” My parents, immigrants from Hong Kong, arrived in St. Louis, Missouri (where I was born) and later moved to sunny Camarillo, California, in the early 1970s. Along the way, I developed fond memories alongside iconic family vehicles, including a 1973 Super Beetle, a 1976 Volvo 240, a 1977 Chevrolet Caprice Classic, and a 1984 VW Vanagon. My father nurtured a passion for sports cars and fast sedans, too. I vividly recall sitting in the package shelf of his 1981 Mazda RX-7 while he shared the wonders of the rotary engine and the thrilling ride in his ‘87 Porsche Turbo. As a Toyota family, I achieved my driving license and, as a reward, my brother’s 1986 Celica GT-S. Six months and three speeding tickets later, I transitioned to a 1983 Toyota 4×4 through college, later graduating from USC with a degree in Biology and a minor in Business. Cars took a back seat during my years as a science teacher for Teach for America, and after exploring freelance photography, I stepped into the editorial world, eventually becoming editor-in-chief of the now-defunct Sport Compact Car. That experience led me to Motor Trend as senior editor in 2007, where I’ve relished every moment.
Share
