It happens to us every day: we try to enter a website and suddenly a grid of poor quality photos requires us to identify all the traffic lights, buses or even fire hydrants even though in Spain, for example, they do not have the characteristic design that is applied in the US. When we solve these puzzles we are not only proving that we are not a robot: we are working for Google.
Google slaves. In the early 2000s, bots were destroying the internet, but a young man named Luis von Ahn had a great idea to stop them. Believe CAPTCHAa system that forced us to identify distorted words to prove that we were human and thus be able to access the content. That system evolved and Google bought the idea and turned it into a perfect system for something we have barely noticed: working for it.
From Google Maps to Waymo. Since then, Google has not stopped taking advantage of the system for two intertwined objectives. The first, effectively, protect us from bots. The second, also known but much juicier for the company, is to turn us all into information taggers. Internet users first recognized words and became a gigantic OCR system that was applied to Google Maps. Then, with images, we ended up helping Google’s image recognition systems improve significantly. That has served, among other things, to feed the Waymo autonomous driving systems.
Statistical consensus. How does Google know that when we choose a fire hydrant or a bus we are responding correctly? It knows this through the so-called “statistical consensus.” Google usually presents images in pairs: one of them (the control image) has already been previously identified by thousands of people, while the other is an “orphan” image that its computer vision algorithms cannot decipher. If you guess the known one, Google assumes you are human and uses your answer about the unknown image to feed its database.
We are the product. All of our readers were probably already very aware of this reality, but now a debate is beginning to activate about the ethics and ownership of digital work. It is something that we already saw with social networks, which were fed by our content, and that certainly also applies to Google: to what extent is it legal for a company to have a huge AI infrastructure thanks to the billions of hours of unpaid “microwork” of its users? Here the famous “If you don’t pay for the product, you are the product.“. It is true that these Google systems have protected us from bots and we have not paid for them “with money”… but with those micro-jobs that we have carried out when solving the puzzles of the reCAPTACHA systems.
Is it possible to poison the algorithm? Here also doubts arise about the true reliability of the system. If a mass group of users decided to mislabel traffic lights or fire hydrants in an organized way, would a self-driving car make dangerous decisions in the real world? That risk seems reasonable, and considering that AI models are increasingly more capable in abstract reasoning and even overcoming captchasan attack by AI bots that did something like this poses a threat worrying.
The invisible CAPTCHA. Google itself knows that visual CAPTCHAs are no longer so insurmountable for machines, so it has been moving its systems towards reCAPTCHA v3a invisible system that does not require you to look for buses, zebra crossings or fire hydrants that you will never see on a street in Malaga or Bilbao. Instead, this system opaquely analyzes your behavior in front of the PC: how you move the mouse, what cookies you have installed and how you navigate. Or what is the same: Google thinks it knows how a human behaves when you’re going to click on “I’m not a robot”… when we’ve been working like robots for years and solving those puzzles.
a brilliant idea. What is clear is that CAPTCHA has been a brilliant idea with implications that not even Google could have anticipated. In fact, it has turned this tool into a way to feed its artificial intelligence systems with our help, without us practically knowing (or caring much). But you know: the next time a website asks you to identify fire hydrants before entering it, remember that you are not demonstrating your humanity. You’re signing on to the afternoon shift at one of the largest data factories on the planet.
GIPHY App Key not set. Please check settings