Following Proco Global’s recent blog series on the impact of autonomous vehicles, we approached Dr. Teresa Escrig to provide expert insight into AI in the automotive industry.
Dr. Teresa Escrig is an artificial intelligence (AI) expert based in Seattle, US, with distinguished accomplishments in research and development of innovative AI products. Her impressive career includes publishing more than one hundred peer-review research articles, four books, four patents and receiving several awards, including the best PhD award and the National Prize on Science and Technology.
Dr Escrig is the founder and CEO of two startups in AI. She currently works as an independent consultant, supporting businesses to develop and implement safe AI roadmaps. Following Proco Global’s recent blog series on the evolution and impact of autonomous vehicles on our Proco Thinking blog, we approached her to provide expert insight into AI generally and also the role that it plays in the automotive industry.
We’re extremely grateful to Dr Escrig for taking the time to answer these questions. This is a rare opportunity to discuss the AI space with an expert with such vast experience and understanding of this field.
Q: What is AI?
The term ‘artificial intelligence’ was coined by John McCarthy the year before the famous Artificial Intelligence Conference at Dartmouth College in the summer of 1956. The purpose of the conference was to create a research field focused on the development of computer programs that emulate people’s behavior or intelligence. Just as there’s no agreement of the definition of human intelligence, there’s no agreement in the definition of AI either.
In general terms, we can describe human intelligence as the ability to communicate, to perceive and understand our environment through vision, smell and touch, to identify and classify things, to make inferences from perceived information and find shortcuts, to find patterns in data to get insights and predict what might happen next, to learn how to do things we didn’t know before, to remember how to do things we’ve previously learned, to glean abstract knowledge from the information perceived, to make decisions based on data or knowledge, to gain wisdom over time, to create art or music, to write a book or a poem, to take enjoyment from a painting or listening to a song, to organize things to create beauty and enjoy that beauty.
Every computer program that emulates one or more aspects of what we consider human intelligence can be called artificial intelligence.
Q: What are the benefits of AI?
According to McKinsey [1], AI has the potential to deliver additional global economic activity of around $13 trillion USD by 2030. The impact of AI might not be linear, but could build at an accelerating pace over time: an S-curve pattern of adoption and absorption.
Potentially, AI might widen gaps between countries. Leading AI countries could capture an additional 20 to 25% in net economic benefits, compared with today, while developing countries might capture only about 5 to 15%. According to Visual Capitalist [2], China might have almost double benefits from AI than the US.
It’s possible that AI technologies could lead to a performance gap between companies that fully absorb AI tools across their enterprises over the next five to seven years—that might see a potential cash-flow growth of about 6% for longer than the next decade—and companies that don’t adopt AI technologies at all, or haven’t fully absorbed them in their enterprises by 2030, might experience around a 20% decline in their cash flow from today’s levels.
Q: What are the main challenges facing mid-market CEOs in the adoption of AI?
Although most organizations have already begun to adopt AI in their business, only 17% of companies have an AI strategy and know how to source the data that will allow AI to work [3], Many organizations still lack the foundational practices to create value from AI at scale.
67% of CEOs think that AI and automation will have a negative impact on stakeholders’ trust.
The AI control resides in data scientists with low or no business knowledge. However, 75 % of AI decisions impact business processes, talent development, the customer experience, corporate governance and ultimately lead to new business opportunities—all CEO concerns.
The fear surrounding the unintended consequences of a largely unproven technology is another main concern: 67% of CEOs think that AI and automation will have a negative impact on stakeholders’ trust.
Another main challenge is the lack of access to the appropriate talent.
Q: What are the main blockers to the uptake and implementation of AI?
One of the main blockers is the lack of transparency, and therefore trust, in machine learning (ML) / black box type of AI. They have been a number of examples of AI going ‘wrong’, including Tay, the Twitter chatbot by Microsoft, using racist language and promoting neo-Nazi views [4] and the bots created by Facebook that were shut down when they were communicating in a language they invented [5].
The industry with the most advanced AI adoption has been autonomous vehicles. The main challenges in this industry include:
- Obscurity of the ML algorithms, even for engineers
- Difficulty of pinpointing errors
- Inability to apply Agile methodologies for algorithm development
- Long periods of testing on the roads with an unproven technology
- Unpredictable time and cost to market
For instance, Uber’s self-driving car struck and killed a woman crossing a roadway in March 2018 because she was apparently considered a false positive [6]. As a result of these challenges, all the players in the industry are scaling back on ambitions, with a tremendous economic impact to these companies.
Q: How does AI benefit the automotive industry specifically?
The most significant use of AI in the automotive industry has been the implementation of self-driving cars. In order to overcome the challenges mentioned in the previous question, it’s necessary to integrate other AI methodologies besides ML, which will bring transparency and repeatability to the algorithm development, as I explained in a previous LinkedIn post
Other applications of AI in the automotive industry include the different advanced driver-assistance systems (ADAS); edge computing, which brings intelligence to the Internet of things (IoT) devices included in the cars; and voice communication to deliver commands to the car, including entertainment.
Q: Do you think the fears around GDPR and negative consequences of AI are justified?
The fear around potential negative consequences of AI are completely justified. That’s why I wrote the book Safe AI.
The greatest danger from AI comes from the technology per se and it has the potential to end humanity. The combination of ML, where the hyperparameters are automatically selected by genetic algorithms (another black box type of AI) and run in a quantum computing with 200k times the computational power of current computers, a narrow goal and no ethics, is something we don’t want to develop.
I also discourage the development of general AI. My proposal is to use holistic AI (integration of non-ML and ML technologies) to enhance humanity’s potential in a way that allows us to increase our intelligence in a slightly exponential way (not only linearly, as it has happened to date) that we access in a device, like our phones, never a chip installed in our bodies because cyber security is a severe, unsolved problem and a chip only facilitates the already extended control over people. If we don’t develop general AI, super-intelligence will never happen. The AI curve will always remain under the human intelligence curve.
The GDPR (European Union General Data Protection Regulation) and the California Data Privacy law were absolutely necessary and force corporations to develop technology that explains the logic of decisions made by AI engines, which will ensure fairness when the decisions affect people or have business implications.
Q: What skills will become increasingly valued in automotive companies with the development and advancement of AI technology?
Real AI experts trained in a broader spectrum of AI methodologies with years of research and development. It will be critical for the automotive industries that want to be at the head of the autonomous vehicle development to work in the integration of non-ML AI and ML methodologies and overcome the challenges currently delaying the success of the self-driving car. Experts with the skills that can bring transparency to AI systems will be critical.
For example, I’m currently incubating a startup, Trace Eye, which provides traceability/ transparency to self-driving cars by showing the relationship between the objects perceived by their cameras and the decisions made by the car (to comply with the mandatory safety regulation ISO 26262).
Interested in delving further into the future of autonomous driving? You can read our recent series of blogs on Proco Thinking: https://www.procoglobal.com/proco-thinking/