CAPTCHA tests are proving themselves to be less annoying, and more helpful in the design and function of driverless cars (Image Source: The Bellingen Shire Courier-Sun)
By Nahum Gale
There is a future where driverless cars cascade down our roads without human interference and minimal traffic, and it would seem we are helping said future come to fruition.
Driverless cars may still be heavily under development, though their success would appear to hinge strongly on the basic human instinct captured in CAPTCHA tests.
CAPTCHA is an acronym for Completely Automated Public Turing Test to Tell Computers and Humans Apart.
The test requires a user to solve a problem, usually consisting of a set of images, that a computer would struggle to autonomously solve.
Most people would be more familiar with its six image test, which asks you to select each one with traffic lights.
But while it may be annoying, there is an underlying reason why we are instructed to take a CAPTCHA test every time we forget our internet passwords, and it is predominantly due to the development of driverless cars.
Raul Barreto is a Senior Lecturer at the School of Economics at the University of Adelaide.
His study into driverless cars was able to shed some light on CAPTCHA’s involvement.
“The idea is for CAPTCHA to be used as a tool to enhance machine learning and thereby apply it directly to driverless cars,” Dr Barreto said.
“This is important because the essence of driverless technology is, in fact, machine learning.
“How does the car distinguish between a dog and a child? Or between a child and a wheelie bin? The way machines learn is through massive repetition.
“If it sees enough dogs crossing the street versus enough children crossing the street, it will eventually be able to discern one from the other.”
Our involvement with CAPTCHA tests essentially assists driverless cars in learning valued judgements that only human consciousness could determine.
However, how does CAPTCHA manage to assemble such vast data to assist our future on the roads?
Zygmunt Szpak is a Postdoctoral Research Fellow at the Australian Institute for Machine Learning Projects, and he was able to address why CAPTCHA is so integral to advancements in driverless cars.
“Driverless cars need to see the world,” Dr Szpak said.
“The best technique for developing advanced algorithms capable of understanding image content involve supervised machine learning.
“Supervised machine learning, however, requires lots of annotated images, including images in which the presence or absence of various objects is recorded.
“CAPTCHA can be used to help crowd-source these required annotations.
“It may be that the CAPTCHA system knows the annotations of a subset of images, and is counting on the user to annotate several additional ones as a side-effect of solving the task.
“The newly annotated images are included in future tests, and if they are repeatedly interpreted in the same manner by many users, they are considered validated.”
So, it would seem, in order for driverless cars to function, they require a human touch: our ability to see the world as more than just a computer screen.
Yet, although it is crucial for these computer bots to adapt our human mannerisms to properly function, it is also important to acknowledge the other functions which are making driverless cars a possibility.
“There are significant engineering and hardware requirements involved in processing the multitude of sensor data that the car acquires in real-time,” Dr Szpak said.
“Consider that a car driving more than a hundred kilometres an hour needs to make a decision and react in a fraction of a second to any sudden change in traffic conditions.
“Autonomous cars also utilise GPS sensors, and 3D reconstructions of landscapes and cities, to localise and orient themselves in the world.”
In short, there are many elements to the development of driverless cars that are assuring these vehicles are a staple of our future.
Still though, it is our own voices that remain as primary ingredients to the assured functioning of these game-changing vehicles.
Like it or not, CAPTCHA is integral to the further advancements of driverless cars.
So, next time your computer asks you to take one of those tests, just remember: you are helping your future artificially intelligent Uber driver not to run a red light.