To have not heard of Chat GPT, Google’s Bard, etc., you must be living under a technological rock. Each such powerful Artificial Intelligence (AI) agent has splashed the news with varying degrees of reception or scorn, e.g., Forbes recently published two articles within five hours of each other entitled “Why AI Must Thrive: Our World Needs It” and “Should We Stop Developing AI For The Good of Humanity.” And for those who have dabbled with the applications, the duplicity of power and danger is nearly inescapable: paragraphs of intelligent repose or responses emerge seemingly instantaneously based upon any simpleton’s prompts. For the masses, though, the practical applications are harder to visualize, let alone imagining the reshaping of an industry. “AI will unleash a new level of productivity and innovation,” predicts Ludovic Hauduc, CTO of Envorso and former VP of Engineering for Meta (in charge of their AI infrastructure). “Just like a large percentage of jobs from the 1950’s no longer exist today, it’s likely that 30% of all current jobs will be reshaped somehow in the next 2-3 years through automation. The speed of innovation around AI is unlike any technological shift we’ve seen before.” And that’s exactly what’s happening in multiple arenas with automotive being no different, specifically with significant changes (and potential pitfalls) in applications, architectures and supplier integrations.https://embedly.forbes.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FeUTYpTdzNhM%3Ffeature%3Doembed&display_name=YouTube&url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DeUTYpTdzNhM&image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FeUTYpTdzNhM%2Fhqdefault.jpg&key=3ce26dc7e3454db5820ba084d28b4935&type=text%2Fhtml&schema=youtube…Insert Text Above
A metaphorical passenger recently commuting on the same Serious-Media-Hype Bus as AI has been autonomous driving, but AI-powered applications run far’n’wide in vehicles beyond just self-driving. These range from enhanced in-vehicle user experiences using more-pleasurable drives (based upon human behavioral sensing) to user-favored music (based upon user and community ratings or skipping) all the way to better chances to survive (given ever-evolving algorithms for sensing the environment and driving patterns). Each application or information “gateway” collects data from on-vehicle sensors, shares it to the cloud, allows centralized servers to improve algorithms for everyone and, simultaneously, uses user-specific data and community-data for tailored-yet-crowdsourced enhancements.
The novices to automotive believe this is a recent transformation. Silicon Valley transformed the world. True, and yet not. Connected vehicles started to ramp-up in the mid-90s, and the adaptation from in-vehicle datahave informed applications such as Predictive Maintenance for 15-20 years.
However, enablers such as improved semiconductors, cellular networks and in-vehicle computing networks have made the flow of information from all parts of the vehicle easier and, therein, almost table stakes for competing. “We are entering a new era,” declared Thomas Weber, Mercedes-Benz’s Development Chief in an interview with Reuters. “Until now, cars retained the properties they had on the day they were purchased.”
To allow for such adaptation, many underlying design changes have and will take place. Processing will be centralized both in-vehicle to allow for lower-cost, efficient edge computing, e.g., move away from 70-80 separate boxes per vehicle to a few super-computers and off-vehicle to allow for improved learning across the entire fleet. Therein, the few, consolidated in-vehicle processors will be designed for extensibility akin to phones, and the computing architecture must allow for future applications to sit atop a platform that can evolve long-past the production date of the metal and gears. To numerically describe that expansion, the 2016 Ford F150 had 150M lines of code, whereas newer vehicles are estimated to reach one billion (1B) lines not including off-vehicle (“cloud”) software. To design for the extensibility of a product that’ll launch in 3-4 years and live in the field for 10-15 years requires a superior, flexible architecture.
“We are the architects of our own operating system, a chip-to-cloud architecture that enables the decoupling of software and hardware,” Mercedes-Benz’s CEO, Ola Källenius explained to investors in April. “We are on a journey to also become a software company. We will put supercomputer-like performance into every single Mercedes.”
Manufacturer strategies like these are a major part of why NVIDIA’s stock exploded 30% last week with an additional 38% growth expected.
Per Källenius’s words, the automotive industry is rapidly changing behind what’s normally mentioned in the press about electrification and automation: it’s softwarification (pronounced soft-WAR-if-ah-kay-shun). Manufacturers have realized that commercial differentiation, complex integration and data security require them to actively control a larger percentage of the exploding software both in-and-out of vehicle. Sometimes that’s by purchasing software — akin to buying the production tooling for hardware — or sometimes it’s as extreme as buying-up companies for their software development capabilities, e.g., GM buying Cruise Automation in 2016, Ford acquiring the BlackBerry team in 2017, Stellantis to spend upwards of $34B on growth including the 2022 purchase of aiMotive.
Regardless of organic or inorganic growth, this shift in the industry will require new capabilities historically difficult for most manufacturers: co-development between OEM and supplier. Since the supplier must provide super-computers with sufficient processing power, they must develop the non-differentiating, base software and integrate the manufacturer’s code thereafter, which has traditionally caused integration quality and project coordination issues. “Finding automated ways to share and continuously integrate code requires coordination from both parties,” suggests Envorso’s CEO and former head of Lincoln, Scott Tobin. “That allows both companies to understand the ongoing truth with transparency from beginning to end.”
And without such transparency, the intelligence that’s artificial will be “who’s winning the race.”