Hell hath no fury like tech bros being told that their self-driving cars, artificial intelligence, and machine-learning technologies are overhyped and nowhere near ready for market. But with that in mind, SFist will hereby enrage the Tesla fanboy, chatbot fetishist, and robot-prepared quinoa loving communities with a peer-reviewed study from the Cornell University Library showing that our new robot-car overlords can be easily fooled or misdirected with so much as stickers, electrical tape, or spray paint being applied to road signs.

Here we see some simple stop-sign defacings that researchers found could easily and reliably dupe self-driving cars into speeding up instead of stopping. “Researchers say that the first image on the left, the one with the ‘love’ and ‘hate’ words fooled a self-driving car's machine learning system into misclassifying the classic ‘Stop’ sign as a ‘Speed Limit 45’ sign in 100 percent of cases,” according to the InfoSec blog Bleeping Computer.

Sure, a blog called Bleeping Computer is not a peer-reviewed scientific journal. But the full peer-reviewed analysis Robust Physical-World Attacks in Machine Learning Models is available online with its rigorous methods detailed, using sentences like “Deep neural network-based classifiers are known to be vulnerable to adversarial examples that can fool them into misclassifying their input through the addition of small-magnitude perturbations.”


It’s time for some game theory plain English. While the ‘love/hate’ defaced Stop sign proved 100 percent effective at fooling autonomous vehicles, other methods also repeatedly duped the robot cars. Graffiti and printed overlays like the ones seen above were able to confuse the robots. That blotchy Stop sign at left was read as a Speed Limit 45 sign in 67 percent of the test cases. That altered Right Turn sign was registered as a Stop sign in 100 percent of the test cases.

The study does note that there are some exceedingly simple fixes to these problems. Road signs could just be made with anti-adhesive or anti-graffiti materials (though this would require a significant infrastructure investment). The larger issue here is that the Uber, Waymo, and other autonomous vehicle pioneers may well be looking past public safety concerns in their desire to see splashy headlines, happy investors, and a return on their vested stock shares. Yes, self-driving cars will inevitably be a reality. But we might want to hit the brakes on their rollout until we fully comprehend their vulnerability to havoc easily wreaked by kids with laser pointers, billboard liberation pranksters or — worst of all — homegrown ISIS operatives.

Related: Security Robot Knocks Toddler To The Ground Then Runs Him Over

Image: Bleeping Computer via Cornell Library