Autonomous cars, what do we have to figure out before we make it happen
Big data and advanced AI for safe self-driving vehicle
Self-driving cars, the proper ones, the ones that will take you to work while you nap/watch the news sipping your coffee, have been part of the collective imagination for quite a while. And it will probably take a long time until someone will be able to choose whether to stay at the wheel or engage in other activities (please, don’t text and drive).
As the progress is evident, and fully automated vehicles seem to be working pretty well in safe environments (suburban areas, dry conditions, basic road layout), some work still has to be done, and the scenario we were depicting just before is most likely not just around the corner. Waymo, Google’s self-driving cars company, is already delivering some autonomous drive experience. Still, it relies on humans supervising the rides remotely, ready to intervene if something unexpected happens. Driverless cars ready to be sold and carry us wherever we want, whenever we want, can be several decades far away.
Large amount of data, efficiency at processing them
Vehicles and infrastructures should be connected together, and cars should be able to tell other vehicles where they are and where they are going.
The conversation about self-driving cars is revolving around some important moral dilemmas, as the AI’s decision making rules in case of emergency (see the good old Trolley Problem).
Whatever the guidelines will be, the possibility to put them into practice depends on the cars’ degree of efficiency in analyzing big data. The further progress in IoT development will play a very important role. Autonomous cars are, in fact, equipped with all kinds of sensors and in order for them to do the trick, namely, to make the cars’ AI more reliable than human drivers, it is necessary that the system has access to an accurate stream of data, and that it has the power to process them in real time.
Humans have biologically learned (at some extent) how to avoid accidents during the painful process of evolution. That’s what we call instinct, or intuition: we have a gut feeling about which plants are going to eat and which ones might be dangerous, for example.
It is a feeling that arises when we must make fast decisions in uncertain situations. We do not have to go through complex reasoning and calculations: we, or our ancestors, have been exposed to certain dilemmas before, and our brain has learned how to react quickly to maximize the chances of success, which means, sometimes, survival.
Machine learning, a sub-field of AI, was developed to teach machines how to do what comes natural to humans. Artificial intuition can give cars the power to immediately and accurately recognize patterns and react, guaranteeing a high level of safety for passengers.
Human beings have a certain sense of the situation on the road, and a concern about their own safety and other people’s one.
We do not need to teach cars how to feel like humans, from the phenomenal perspective, but how to behave like they do.
The future of autonomous cars depends on our ability to build for them a “nervous system” that will guarantee us, as much as possible, that they will make decisions which are compatible with our moral sense, ideally, with a higher speed of reaction and a better accuracy.
How do we achieve our goal? With powerful IoT (and sensors) and AI\Machine Learning technologies. The ability to exchange a high amount of data, extract relevant information, and know what to do with it.
The Automotive developers who will be successful and quick at achieving these safety goals will win the race, and finally allow us to play our favorite video games while stuck in the traffic on the highway.