Before you go...
Did you enjoy the reading? Did you notice this content is free and also ad-free?
We intend to keep it this way so please subscribe to our newsletter so we can stay in touch.
Don't forget to share it with your friends so they can enjoy the reading too.
Join Our Newsletter!
* Please provide a name
* Please enter a valid email address

The Autonomous Vehicle Hype

What happened to the bold predictions and big investments behind driverless cars?

By Terrance J. Mintner • August 14, 2020

Bram Van Oost | Unsplash

About three or four years ago the tech world was abuzz with talk of autonomous vehicles (AVs). Tech gurus chatted excitedly about how the sector was quickly advancing. Some analysts and CEOs boldly predicted that in just a few years these cars would be a common sight on the roads.

Well, that didn’t happen. Was it all hype? A big ball of nothing?

It turns out the obstacles – both technological and regulatory – were bigger and more numerous than expected.

And then there’s COVID-19, which has hurt manufacturers. According to a recent survey by McKinsey, car sales have plummeted “71% in China in February 2020, 47% in the US in April, and 80% in Europe also in April.”

But there are bright spots.

“Companies still invest a lot of effort and money into autonomous driving,” Nitsan Avivi, Tech Scout for Konnect, the innovation hub of Volkswagen Group in Israel, tells Spectory. “For example, Volkswagen just invested $2.6 billion in Argo AI as part of its cooperation with Ford.”

Public Trust

Autonomous driving, Avivi continues, “is still the end goal of every manufacturer in the world, but first of all there is the issue of acceptance and trust among the public. You need to have people convinced that AVs are safe enough and there is no threat to their personal data and privacy.”

Ben Voklow, Founder and CEO of Otonomo, an automotive data services platform, tells Spectory that “automotive industry leaders need to build trust and awareness of AV technologies among consumers to aid their mass adoption.

“Additionally, governments need to take action to support the integration of AVs, from designated AV lanes to tax incentives for corporations to deploy AV fleets. Working together, the auto industry and civic leaders can smooth the way for progress in the AV arena.”

Regulatory Obstacles

On the regulatory end, Avivi explains that for AVs to become a reality, authorities must allow them on the roads. But not only that, they must also clearly define the liability in case of an accident.

“Think about it. If you’re driving your car and crash into a wall, it’s your problem, unless you can prove that there was a fundamental flaw in the car and the manufacturer knew about it and didn’t fix it.

“But if an autonomous vehicle crashes into a truck – and we’ve seen this happen with Teslas in semi-autonomous mode more than once – the driver or his/her relatives or those affected can sue the car company. Such a lawsuit, especially in the US, can end in hundreds of millions of dollars in compensation. No car company could withstand this.”

Slava Bronfman, CEO of Cybellum, an automotive cybersecurity company, tells Spectory that “the car insurance business model today is built around the driver as the entity liable for the car and its behavior on the road.”

But, he explains, this model is irrelevant for AVs.

“Just think of an AV hitting a pedestrian because of an incorrect algorithm judgment. Who is liable for that? Is it the driver, the car manufacturer, the dealership that sold it or maybe even the municipality in which the accident took place, since it is responsible for the infrastructure?

“This is a big open question, which must be addressed prior to seeing a massive rollout of autonomous cars on the road,” Bronfman adds.

Denys Nevozhai | Unsplash

The Current Technology

Let’s dive further into the technological issues. Where does the current technology stand and what challenges lie beyond its grasp?

First off, Avivi explains, there are five levels of AVs. It goes from Level 1, a car with barely any autonomous capacities, to Level 5 which is a car that doesn’t even have a steering wheel. “You just get in and say ‘take me home,’ and it drives you home.”

“In between you have different levels of autonomy. Level 3 means that if you’re driving in very specific conditions – usually involving good weather and visibility, open roads, on the highway and not in the city – you can let the car drive itself but you need to stay alert and be ready to take control of the car again in seconds.”

“This is still autonomous driving. So when people said we would have autonomous vehicles by 2020 or 2021, it seems they weren’t wrong, at least not by much.”

Just think of an autonomous vehicle hitting a pedestrian because of an incorrect algorithm judgment. Who is liable for that?

Bronfman adds that “there is still a lot of research and development that needs to be done before we’ll see Level 4 or Level 5 vehicles on the road.

“It’s very challenging to develop autonomous driving systems that cope with infrastructure that was originally planned and built in an era without computer vision and AI, and was designed for human drivers and pedestrians.

“Simulating, testing and validating all edge cases is an enormous task that has yet to be completed.”


Now that cars are connected to the internet while on the road, they are at risk of being hacked. The more connected each component is within the car, the higher the risk it could be compromised by hackers.

“As many of the car’s systems are software-based and as the car as a whole is connected to external networks, the car’s attack surface is growing exponentially,” Bronfman says.

“For example, infotainment systems and telematics units are typically connected to networks using GSM, Wi-Fi and Bluetooth technologies, which provide a stable connection that hackers can use to penetrate the car. Hackers can access Personal Identifiable Information (PII) such as personal contacts, navigation destinations or even a home address and mobile phone information stored in the vehicle.

“So, the need to protect the cyber posture of the car is clear, both on the operational side and with regard to privacy.”

Advanced Testing

Roy Fridman, Vice President of Business Development and Sales at Foretellix, a company that has developed intelligent automation and analytic software tools to ensure the safety of automated driving systems, tells Spectory that “the largest barrier to broad deployment of autonomous vehicles is undoubtedly safety.”

“How do I make sure the AV is performing well under hundreds of millions and billions of scenarios and edge cases? How do I provide a quantifiable benchmark to say that I have tested enough? Companies must be able to provide good answers to these questions before we see AVs at broad deployment.”

This will require, Fridman explains, a shift in mindset, as well as advanced automation testing, and new methodologies on how to manage the complexity of testing in an almost infinite space of possibilities.

Artificial Intelligence

Where does AI factor into all of this?

David Barzilai, Executive Chairman and Co-founder at Karamba Security, a company that provides embedded cybersecurity solutions for connected systems, believes that manufacturers initially miscalculated how complex the technologies are for AVs at levels 4 and 5 – especially computer vision and AI.

“This barrier entailed significantly more investments than planned,” he tells Spectory.

One of the big challenges of getting these vehicles on the road is “computer vision and analysis capabilities, which enable the car to detect and overcome obstacles that drivers overcome in the manner of common driving practices.”

Barak Matzkevich, COO of Cartica AI, a company that has developed an automotive visual intelligence platform, tells Spectory that “the capability to perceive every possible scenario out there – the current approach in deep learning – is coming to a stretch.

“In deep learning whatever you train the system for, hopefully, when it comes across certain scenarios out there, they were covered by the training,” he adds.

But to unlock further advances, Matzkevich explains, there needs to be a different approach with AI. This is because one can never know what edge cases are out there or what form various scenarios will take.

“It is fine to cover 99.9% of them but once you are out there driving you cannot afford that 0.1% – that the system will not be able to perceive or detect a pedestrian or a cyclist.

“What is the way to overcome that barrier? The answer is a machine that will be able to understand, overcome and learn basically, even if it encounters for the first time a certain situation involving things that were not originally included in the training session,” he concludes.

The challenges outlined above are certainly formidable, but let’s end on a positive note. Why not?

Investments. They are not getting smaller. To the contrary.

According to a recent article by Forbes companies are entering into strategic partnerships and making acquisitions to give them an edge in the AV race, even during COVID-19.

We’ve already mentioned Volkswagen. The company has pledged to invest about $71 billion for what it calls the “car of the future.” This includes a good chunk of funding for AVs.

Other companies are investing in AV development as well: General Motors $20 billion, Ford $4 billion, and Amazon $1 billion.

Not too shabby.

Ashley Harkness| Unsplash

Leave a Reply