Recently, the world was shocked when Uber’s self-driving vehicle was blamed for the death of a pedestrian. This harked back to Isaac Asimov’s first law of robotics stating, ‘A robot may not injure a human being or, through inaction, allow a human being to come to harm’. This tragedy caused a halt in the fervour to bring about this technology to the public as several companies have frozen their pursuit of self-driving car research.
But his hiatus won’t last. Self-driving cars have the potential to change the social fabric of society in a multitude of ways.
How does autonomous driving work? Why couldn’t we do this before? How will society change when it becomes mainstream?
Autonomous driving works by combining a few different technologies that, until recently, were something of science fiction. One of the main technologies is something called “Deep Learning”. Deep learning is an aspect of artificial intelligence used to recognise patterns in data, including pictures. You’ve probably already been astonished by deep learning without even knowing what it was called. Ever wondered how social media sites can tag you in a group photo taken by someone you’re not connected with? Deep learning.
The cameras of self-driving cars are making decisions based on the exact same technology – sounds way crazy, I know. A computer program looking at a picture and making decisions in real-time. Just brilliant!
This technology isn’t relegated only to large companies employing scores PhDs. If you create an account on Amazon Web Services you will see a wide array of Deep Learning Amazon Machine Images. There are even free instances you could use that are loaded with the exact same tools that Google uses to identify pictures and guess you’re about to type: ‘How do self-driving cars work’?
Deep learning software requires lots of computing power. What powers these complex computations? Mostly graphics cards. The same graphics cards people are using to play intricate video games and mine cryptocurrencies. The computational speed in which this is done was impossible just a few years ago. There have recently been several published benchmarking tests showing how much faster graphical processing units (GPUs) can train a neural network using TensorFlow (one of the deep learning software packages freely available on AWS instances). In many cases, GPUs were 10x the speed of current general-purpose processors - such as those found in new, high-end laptops and desktops - for this particular task.
Faster, more robust mobile networks
The second technology that is going to make self-driving cars more feasible is 5G mobile networks. Now, everyone is aware of 4G LTE technology. It was introduced in 2010 and significantly increased data access rates compared to its predecessor, 3G.
So, what does 5G technology bring? Very high speed mobile data. Up to 400 Mbps (megabits per second) in ideal circumstances. Is that a lot? It doesn’t sound like a lot considering my laptop has more than a terabyte of disk space. Then, you realise that you can comfortably stream 4K Ultra High Definition video content at about 25 MBPS.
5G is important for ‘The Internet of Things’ which will include self-driving cars. We probably won’t see 5G rolled out until at least 2019, but it’s going change our lives by enabling massive data to traverse mobile network as if the devices were connected to fibre optic cables.
What are some of the necessary 5G actions that are required by autonomous cars?
Synchronising with traffic lights for starters. Imagine the cost savings on energy if the billion drivers world-wide were never surprised by a yellow light and had to brake suddenly. I’m certain this would prevent an uncountable number of accidents as well. Also, real-time updates on traffic conditions to optimise routes driven and communicating with adjacent vehicles to coordinate lane changes or exiting the freeway safely. Imagine if every car around you on the freeway had a 60-second ‘heads up’ that the next exit was yours and they had the ability to coordinate their adjustments to let you safely exit. This would be the antithesis of ‘road rage’.
The testing and verification of 5G mobile technology is paramount to ensuring safe autonomous driving. Our society is becoming increasingly dependent on the fidelity of our communication networks and this will hold especially true for 5G.
Security of self-driving cars
Skilled hackers, hijacking the controls of self-driving cars, and delivering your favourite celebrity to your BBQ would be a news-worthy prank. Not that long ago, the security community watched Charlie Miller commandeer the controls of a moving jeep while journalist Andy Greenberg sat inside, documenting the experience.
The more control over our lives we give to technology, the more we open ourselves to abuses of that technology. This is why thorough security testing and skillsets involving secure code writing are becoming more important than ever. If the world was outraged when Equifax lost our personal credit information, can you imagine the response if it were a life-or-death scenario? Or the impact to global commerce if a nation-state could disable the vehicles of a country remotely?
Road rage would take on a whole new meaning if a frazzled hacker had the ability to swerve a vehicle into danger remotely. These are the things we need to prepare for and do so with the seriousness it deserves. This approach is being pioneered by OpenAI whose mantra the discovering and enacting the path to safe artificial general intelligence. Government regulation over this industry, including mandatory security auditing of all networked software in vehicles, will be compulsory before we surrender control of our vehicles.
Social impact of self-driving cars
How will our lives change when our cars are self-driving? I believe this technology goes far beyond a simple convenience, like auto-correcting text messages.
Self-driving cars will change social behaviour in a significant way. The interior of vehicles will be crafted to maximise office productivity and for entertainment. Binge-watching media on a large screen or video conferencing in your car will be the norm. Taking a nap in traffic, when you’re alone in the car, will be a sweet respite from the busy world. Even the busy streets of Mumbai will be absent of incessant honking as the former drivers recline while reading the news or communicating with friends and colleagues.
Concerning our initial subject of Uber’s self-driving car hitting a pedestrian, reports show that there were over 1 million traffic fatalities globally where there were people behind the wheel. If actuarial math can estimate that fatalities will be drastically reduced by self-driving cars, will society demand their ubiquity on the roads? Or will society be unable to stomach a deep learning algorithm making decisions that could result in loss of life when an algorithm determines one passenger’s life should be sacrificed in order to save two pedestrians?
These are the choices we must make as a society and we’re going to need to make them sooner than we think.
Eventually, people may not even learn how to drive. It will have gone the way of composing a formal letter by hand. This could end millions of jobs in the line of trucking and delivery. This trend has been ongoing since the industrial revolution where job destruction takes place due to advancements in technology. I am confident that we are going to continue to adapt because technology has always created more jobs than it has destroyed.
The nostalgic idea of the ‘road trip’ will burst back into fashion. A family of five will weigh the cost of plane tickets against jumping in the self-driving car after dinner and falling asleep after a movie to wake near their destination the following morning. No need for full body scans to enter the airport terminal. Your bottle of water is more than welcome in your car. And you dictate the moment of departure.
Energy efficiency is another macroeconomic factor in favour of this technology. Self-driving cars will travel at optimal speeds to maximise energy efficiency and take the most expedient routes. Even a small energy savings, multiplied by a billion daily car trips, would be significant.
Also, will this technology be contingent on electric cars? Will that require massive changes to the power grid? Will driving increase significantly now that people feel comfortable working farther away from urban areas now that their cars are mobile office? Or will private car ownership decline and we simply summon driverless vehicles for each trip via our devices.
All of this will unfold in our lifetimes.
Paradigm shifting technology, such as allowing artificial intelligence to become humanity’s chauffer, requires prudent steps to be taken. Testing the security of these emerging technologies, as well as the mobile networks that they leverage, is of the utmost importance. We all want to enjoy lowered blood-pressure and a binge-worthy show while deep learning and 5G mobile schlep us to and from our, already stressful enough, jobs.
Phil Trainor, Principal Product Manager, Applications and Security, Keysight Technologies
Image Credit: Karsten Neglia / Shutterstock