Stop anyone in the street and ask them whether they’ve ever encountered artificial intelligence, and they’ll probably cite an entertaining experience goading a voice assistant such as Alexa or Siri. Some might mention a frustrating experience with telephone call centre automation, others an online ‘help’ chatbot. The fact that many customers would mention these, where once they would probably only have referenced a sci-fi movie or two, underlines how far into our lives AI has already advanced.
But what relevance does Alexa’s ability to talk back have for the automotive industry? And why would car manufacturers be interested?
BMW, Daimler, Toyota, Hyundai and Kia are just some of the vehicle manufacturers working on AI, initially for in-vehicle infotainment and ultimately for autonomous drive applications. So, too, is the Volkswagen Group. But whilst product is integral to a car manufacturer’s business, there’s so much more that such a company can do with AI.
PwC defines AI as ‘a collective term for computer systems that can sense their environment, think, learn, and take action in response to what they’re sensing and their objectives.’ Deploying AI in corporate processes, large organisations can speed up data analysis and the calculations required for forecasting, increase efficiency, and free up employees’ time to focus on tasks where they can add real value.
Welcome to the machine
The secret lies in the algorithms that facilitate machine learning, says Dr Martin Hofmann, Chief Information Officer at Volkswagen Group. “Machine learning algorithms process vast quantities of data by detecting patterns. Based on these patterns, and changes in these patterns, the algorithms can make independent predictions. And based on these predictions, the machines can make decisions.”
Computers have for decades crunched numbers to produce outcomes beyond the capabilities of human calculation, but the rate at which this can now be done is such that IBM’s Watson is able to read millions of pages per second. Apply this to corporate data, and it’s easy to see how a business could operate far more efficiently. And it is here where a major organisation such as Volkswagen stands to reap the commercial benefits of researching, adopting and ultimately deploying AI in key areas of the business.
AI could boost global GDP by 14% by 2030, believes PwC, adding US$15.7trn to the global economy. That figure, the company notes, is greater than the current output of China and India combined, and includes a US$6.6trn contribution from increased productivity.
“For us, the key question is, can we augment the human being by giving that person added artificial intelligence?” says Hofmann. “Consider, for example, the laborious data analysis required to predict vehicle sales, including all the different components in all of the different countries in which we sell. This is a multi-dimensional problem, and today this is done with Excel sheets and human brains,” he notes. “Now think of the complexity of our business: we’re in 120 countries, we sell 10 million vehicles a year, and we make 250 different models. It’s nearly impossible to accurately predict what we are going to sell in a specific market for a specific configuration.”
In this example, a machine learning algorithm would process the data, detect patterns, and then make predictions or recommendations. “Our human sales planners would receive pre-defined scenarios recommended by the AI engine,” explains Hofmann, “and they can then consider scenario A, B or C, and choose whether or not to follow that recommendation. AI would be making life easier and enabling humans to handle complexity.”
Servers processing data; it’s much less glamorous than the popular concept of human-like robots with human-like perception. It also underlines how machine learning for enterprise robotics differs from manufacturing robotics. Robots are machines programmed to follow and endlessly repeat a sequence of instructions. Machine learning algorithms learn on their own, interpret and ‘understand’ the data they are provided.
As well as improving corporate and strategic processes, VW Group sees the potential for AI in manufacturing. Production management loves automation – it increases speed, efficiency, accuracy, quality and safety (just ask the fingerless metal press operators of as recently as three decades or so ago). But classic automation can only do so much – repetition and strict adherence to specific lines of code can be limiting. Automation for the next generation needs to be smart, and that’s where AI comes in.
According to a 2017 discussion paper published by McKinsey Global Institute, “Advances in AI technologies will enable the industry to leverage rapid growth in the volume of data to optimize processes in real time. They can shorten development cycles, improve engineering efficiency, prevent faults, increase safety by automating risky activities, reduce inventory costs with better supply and demand planning, and increase revenue with better sales lead identification and price optimization.” With the addition of AI, suggests McKinsey, manufacturing can be smarter, more nimble, and less prone to error.
“Put simply, we want to make robots intelligent so that they don’t need permanent human input to do their task,” says Hofmann. “In return, they will take over many of the tasks that human beings do today that we perceive to be repetitive or stressful. Ergonomics is a big issue for us, and we’re always looking for robots that can take over certain heavy duty, physically demanding and repetitive tasks that humans shouldn’t do.” An AI-enabled robot could learn to accommodate anomalies, such as identifying the need to adjust its movements for a misplaced component.
Thus, another area being explored by VW Group is human robotic interaction, or HRI. “Volkswagen is the first company to have developed learning technology algorithms where the robots identify and learn human movement and human intention in a manufacturing setting,” explains Hofmann. “A robot can hand over components to a human operator and, depending on how they move, it might slow down or move differently. And that’s done by sensing and learning human behaviour. When another worker comes along, such as at shift change, the robot will adjust to that person. This whole HRI concept allows us to make robotics much more flexible and to let humans collaborate more easily with machines.”
The idea of collaborative robots in manufacturing is not new, but it’s still too early to know whether these cobots are an accepted feature on the factory floor. “It will happen in the near future,” assures Hofmann. “We’re piloting it in a live environment, learning from it, and then we will scale it.” Crucially, however, these intelligent cobots would only be deployed to help – not replace – human operators. “AI must always help humans in a meaningful way. Intelligent robots will learn to optimise themselves, but always to support the human being. In our body shop, we have a high degree of automation. with robots, but now we’re talking about using human and robot collaboration for the assembly of very critical vehicle components, for example. It’s a true collaboration,” he insists. According to PwC, labour productivity improvements are expected to account for over 55% of all GDP gains from AI over the period 2017-2030.
To pick up on an earlier point, AI is increasingly being used in call centres and on customer service chat websites. It’s clever, but it’s often clunky, and consumers quickly see through the pretence of human interaction. What implications does this have for the AI in production line robots that have to make decisions based on how the operative next to them is moving in order to complete a complex procedure?
It’s very different, assures Hofmann. “In a call centre, customers could present a million different scenarios. In the case of manufacturing robotics, the AI operates in a professional environment where the number of tasks is limited and clearly defined,” he explains. “Here, the robot has a specific task, and to accomplish this task the robot learns how best to do it. You can say anything you want to Google Home or Alexa, so their algorithms need to cope with many different variations of possibilities. There, the technology is not as advanced as it should be. But for a specified range of functions, such as a factory setting, it works really well.”
One of the challenges for a traditional organisation as large as Volkswagen is to ensure that, when taking on an AI-sized challenge, it has the capabilities to operate at the bleeding edge of the technology.
In 2014, the company established its Data:Lab in Munich, specifically to specialise in AI, automotive data science and machine learning. It was created with investment from Volkswagen Innovation Fund, a joint initiative of VW and the Works Council which supports the company’s initiatives beyond its core business areas. The primary aim is to improve efficiency and accelerate corporate procedures by identifying those processes best suited for AI. It also works on a number of other areas, ranging from cyber security to traffic flow optimisation.
“It’s probably the biggest accumulation of AI talent in the automotive industry in Europe right now,” says Hofmann. Headed up by Patrick van der Smagt, a professor from Technical University of Munich, a 70-strong team of physicists, computing linguists, AI experts, robotic experts and mathematicians works on machine learning algorithms that can identify and predict patterns. This includes a small team which conducts fundamental research, publishes papers, liaises with academia, and files patents. “We also support many universities around Europe, and collaborate with partners from the tech industry. And a second team in San Francisco keeps us in the loop,” Hofmann adds.
“The AI research team is tasked with looking five to seven years ahead. The other 60 people work on applied AI, tasked with implementing the algorithms throughout the corporation. And our 12 brands develop and implement these algorithms in the different locations in the world where we have operations.”
There’s a difference between the development of AI for business purposes and the development of AI for autonomous driving – and while the researchers do work together, “we keep them separate,” says Hofmann. “The autonomous vehicle group focuses on electronics, and picture pattern recognition. They are managed by Audi, which is responsible for autonomous driving development within the VW Group. Those researchers are located in Munich, in the same building. They exchange people and projects, but these are two separate organisations.”
A major area of concern for the development of AI is, unsurprisingly, the very real threat of malicious outside intervention. VW, however, sees hacking, data theft and cyber crime not just as a threat to AI, but also an opportunity, something that AI can be used to tackle. “It’s a critical issue,” agrees Hofmann. “Just as with any technology, there’s a fine line between use and misuse. The Internet is one of man’s biggest technological advances, yet the Dark Web presents a serious challenge. I think AI is a tremendous opportunity to make life better and more comfortable, and make driving safer. But we are not naïve –there will be forces out there misusing that same technology.”
Putting the AI into creativity
Readers will no doubt be familiar with the Turing Test, developed in 1950 by Alan Turing to assess a machine’s ability to convince someone that they are interacting with another human rather than a machine. Advances in AI suggest that it is no longer a question of if, but when, AI can convincingly pass the Turing Test, and thus begin to threaten those in executive positions, assembly line roles or even, perhaps, journalism. Time will tell whether human creativity is something that can ever be achieved by AI.
An AI robot might be conducting this interview a few years from now, chuckles Hofmann. “It’s a good point, and yes, it is a question of creativity, which is based on experience. But experience is nothing more than data interpreted, and machines are outpacing everything we know. In the long term, there will be a coexistence of algorithms and machines that do specific tasks better than humans do today. But the moment there is no data available, or a prediction cannot be made, that’s where human creativity comes in – with the possibility of failure.” AI should not fail, and algorithms are written such that the failure rate is minimised to near-zero, he points out. “But humans do fail, something which we accept. At the end of the day, at least in a decision-making process, the final call should, in critical cases, lie with a human being. That is an ethical discussion which will run and run.”
The notion that AI might threaten jobs is particularly pertinent to a company such as Volkswagen, a traditional and leading German company, headquartered in Germany with a powerful Works Council. It would be reasonable to expect the Volkswagen workforce to be concerned and suspicious about the development of AI, and the implications for the company’s 640,000-strong workforce. Were it not for the aforementioned involvement of Volkswagen Innovation Fund, Hofmann’s response might come as a surprise.
“From a very early stage, our employee representatives in the unions became involved, and they are co-driving this,” he reveals. “They know that if we don’t apply new technologies, we would be at a competitive disadvantage. But we are using AI and applying it in a way that is helpful and respectful to the company, and the collaboration with our unions and employee representatives is very strong.”
The typical consumer referred to at the outset will, for some time yet, think of AI as an advanced voice-controlled jukebox, or a frustratingly inflexible call-centre automaton; a typical VW Group assembly line operator, however, might soon see AI as something that makes their job easier, and those teams in production forecasting or logistics planning might find the data they work with more detailed than before and surprisingly accurate.
For the VW Group, AI is a key competitive factor for future business; it’s not an end, but a means to an end. And for Hofmann, the choice is simple: develop artificial intelligence in-house, or prepare to depend on the intelligence of others.
This article appeared in the Q1 2018 issue of Automotive Megatrends Magazine. Follow this link to download the full issue