Finding the ‘blind spots’ in autonomous vehicle artificial intelligence

Autonomous vehicles are becoming more and more sophisticated, but concerns still abound about the safety of such systems. Creating an autonomous system that drives safely in laboratory conditions is one thing, but being confident in those systems’ ability to navigate in the real world is quite another.

Researchers from Massachusetts Institute of Technology (MIT) have been working on just this problem, looking at the differences between how autonomous systems learn in training and the issues that arise in the real world. They have created a model of situations in which what an autonomous system has learned does not match actual events that occur on the road.

An example the researchers give is understanding the difference between a large white car and an ambulance. If an autonomous car has not been trained on or does not have the sensors to differentiate between these two types of vehicle, then the car may not know that it should slow down and pull over when an ambulance approaches. The researchers describe these kind of scenarios as “blind spots” in training.

blind spots autonomous vehicles mit ai spot 0 1
A model by MIT and Microsoft researchers identifies instances where autonomous cars have “learned” from training examples that don’t match what’s actually happening on the road, which can be used to identify which learned actions could cause real-world errors. MIT News

To identify these blind spots, the researchers used human input to oversee an artificial intelligence (A.I.) as it goes through simulation training, and to give feedback on any mistakes that the system makes. The human feedback can then be compared with the A.I. training data to identify any situations where the A.I. needs more or better information to make safe and correct choices.

“The model helps autonomous systems better know what they don’t know,” author of the paper Ramya Ramakrishnan, a graduate student in the Computer Science and Artificial Intelligence Laboratory, said in a statement. “Many times, when these systems are deployed, their trained simulations don’t match the real-world setting [and] they could make mistakes, such as getting into accidents. The idea is to use humans to bridge that gap between simulation and the real world, in a safe way, so we can reduce some of those errors.”

This can also work in real-time, with a person in the driving seat of an autonomous vehicle. As long as the A.I. is maneuvering the car correctly, the person does nothing, but if they spot a mistake then they can take the wheel, indicating to the system that there is something that it missed. This teaches the A.I. in which situations there are conflicts between how it expects to behave and what a human driver deems safe and responsible driving.

Currently the system has only been tested in virtual video game environments, so the next step is to take the system on the road and test it in real vehicles.

Emerging Tech

Awesome Tech You Can’t Buy Yet: Grow veggies indoors and shower more efficiently

Check out our roundup of the best new crowdfunding projects and product announcements that hit the web this week. You may not be able to buy this stuff yet, but it sure is fun to gawk!

Waymo rules and Apple trails in California self-driving car benchmarks

California's DMV releases annual reports of self-driving car disengagements on public roads. In the most recent reports. Waymo had the best performance, GM Cruise came in second, and Apple's self-driving program was in last place.

Researchers teach self-driving cars to predict pedestrians’ next moves

University of Michigan researchers are developing a system that teaches self-driving cars to predict pedestrian movement. Humans don't always act in their own best self-interest, so autonomous cars will need to practice protective driving.
Emerging Tech

Statistician raises red flag about reliability of machine learning techniques

Machine learning is everywhere in science and technology. But how reliable are these techniques really? A statistician argues that questions of accuracy and reproducibility of machine learning have not been fully addressed.

2019 RAM 1500 Classic Warlock special edition: Badass style without the whoop

If you like the looks of blacked-out badass trucks without the cost of a desert racer, FCA announced the 2019 Ram 1500 Classic Warlock, a special edition pickup that focuses on appearance with only a touch of additional off-road capability.
Product Review

Mercedes-Benz updates the timeless G-wagen for the modern world

For decades, the G-Class has been an outlier in the Mercedes-Benz portfolio, a body-on-frame brute with the soul – and driving manners – of an off-road pickup. With the all-new G550, Mercedes seeks to smooth out some of the rough edges.

2020 GMC Acadia toughens up on the outside, gets smarter on the inside

The 2020 GMC Acadia crossover gets styling updates and a more rugged AT4 trim level. Under the skin, the Acadia sports a new turbocharged four-cylinder engine and a wider array of tech features.

Nissan is using old Leaf batteries to power and connect off-the-grid campers

Nissan has teamed up with trailer manufacturer Opus to design a mobile, weatherproof power pack built with battery cells sourced from the first-generation Leaf. Called Roam, the pack stores enough electricity to power a camper for up to a…

Alfa Romeo’s latest Ferrari-powered F1 race car is ready to hit the track

Alfa Romeo is doubling down on Formula One racing after a decades-long hiatus. Now essentially a support team for Ferrari, its 2019 driver lineup includes a former world champion and a potential future star.

Subaru’s latest VIZIV concept car is pumped full of adrenaline

The Subaru VIZIV Adrenaline is the seventh member of the Japanese automaker's family of VIZIV concept cars. It debuts at the 2019 Geneva Motor Show, but for now, all we're getting is a shadowy teaser image.

Audi is advancing the tech that teaches cars to talk to traffic lights

Audi is teaching its cars the language of traffic lights. The company developed technology that tells motorists what speed they should drive at in order to catch as many green lights as possible.

Watch a modified Audi e-tron electric SUV drive straight up a ski slope

A modified Audi e-tron climbed up an 85-percent gradient on an Austrian ski slope in a tribute to a classic Audi commercial. The vehicle used for the stunt sported an extra electric motor and spiked tires.

Mamma mia! Alfa Romeo will unveil a new model at the Geneva Auto Show

Alfa Romeo told Digital Trends it will unveil a new model at the 2019 Geneva Auto Show. It stopped short of revealing what it has in store, but rumors claim it will be a crossover positioned below the Stelvio.

Citroën says you could drive its tiny Ami One electric car without a license

Citroën's Ami One concept car is an electric vehicle that's as cute as it is compact. The miniature motor only has a top speed of 28 mph, so the French automaker imagines it as a shareable runaround for short drives.