Banner AI

How you and I are teaching Artificial Intelligence how to drive.

Artificial intelligence is one of the ultimate buzzwords at the moment. But what is it exactly and how can it be created? And most importantly, how are we helping it to perform even better? Marvin Minsky, a cognitive scientist, described Artificial Intelligence as the science of making machines do things that would require intelligence if done by humans.

Well, that seems pretty obvious isn’t it? Just like human beings, artificial intelligence shows the ability to use external data and uses this data to reach its goal to the best of its abilities. That’s where one of the issues comes in. Right now, a lot of these AI systems can grab and use this external data, but they struggle to know what is ‘correct’ and how it should interpret this data. It still needs a third party to verify whether or not the AI is on the path to reach its goal.

That’s where you and I come in. The best way to make Artificial Intelligence resemble Human Intelligence, is, well, humans. But if one person had verify everything AI is trying to decipher, he’d have a burn-out in no-time. If millions would verify a solution multiple times, however, it would be way more efficient.

That’s one of the reasons why CAPTCHA was developed. CAPTCHA stands for Completely Automated Public Turing test to tell Computers and Humans Apart. What an acronym. In short it’s a huge test everyone can fill in, to make computers so smart, you couldn’t tell them apart from humans. At  least in intelligence.


Do you remember the tests like the one on the left? I’m sure you’ve filled in several before. This is a great example of a CAPTCHA, made by Google’s reCAPTCHA.

Back in the day, Google was trying to digitise old books, but their scanning intelligence had some problems identifying all of the letters and words. That’s why they took a few excerpts from different books, threw them online to let you decipher them and bang, you just helped making Google’s AI smarter.

A bit later, Google launched their new project, Google Streetview. Another problem arose, because even though they had taken pictures of all the buildings, their intelligence couldn’t read the exact numbers. That’s when everyone suddenly saw images like the ones below popping up. And guess what, everyone who at some point in time typed in the right street address made Google Streetview to what it is now.


The great thing about this insight is that you and I can now reverse-engineer Google’s next steps in AI through their reCAPTCHA strategies. Lately I’ve been bumping into other ones online and they looked a little like the one below. Google is asking us to verify whether or not there are street signs, lampposts and street lights in the picture. We do know that they are testing self driving cars, so I have a pretty valid hunch we are teaching their AI how to drive.


I’m wondering what Google’s next reCAPTCHA will be. Verifying Celebrity faces? Recognising different kinds of foods? Identifying emotions? How do you think we will help Google in the future?

P.S. If you want some more insights on the future of mobility, be sure to download our free Vision Slide Deck through this link:

Geen alternatieve tekst opgegeven voor deze afbeelding

About Maxim Van Meenen

Maxim is a marketer on paper, but a problem solver at heart. With his new age thinking and hunger for trends he wants to show the impact of digital and help businesses find a way to jump these hurdles. His interest for new gatekeepers like the GAAFA players (Google, Amazon, Apple, Facebook and Alibaba) fuels his eager to not merely follow, but to break down and thoroughly analyse the trends these tech players are creating.