Bothersome bots and the birth of Captcha

The world of Captcha (Completely Automated Public Turing test to tell Computers and Humans Apart – yes, it’s a bit of a mouthful) has a relatively short but dynamic history. The idea was introduced by German-Guatemalan entrepreneur Luis von Ahn, also the co-founder of Duolingo, who was a PhD student at Carnegie Mellon in 2000 when he heard about Yahoo’s struggle in dealing with the ever-increasing spam bot accounts. In response, he invented a test to tell robots and humans apart – one that a computer was incapable of passing. Given that humans are naturally skilled at recognising optical characters, the test originally took the form of a warped text.

You may have also noticed, though, that these tests have increased in difficulty over the years – and for good reason. While computers of the early 2000s were incapable of successful optical character recognition, the very data provided by those filling in Captchas quickly made these systems smarter. In 2005, reCaptcha was launched, this time using two words for increased difficulty, and eventually, after computers began outperforming humans at deciphering even the most distorted of characters, it was time for reCaptcha v2. 

Look in the mirror to find AI’s favourite teacher

This is the system as you know it today: one based on images instead of texts. With every fire hydrant and crosswalk you identify, you are training computers to recognise objects in the real world. This has already had exciting, tangible effects, with Google using reCaptcha to make their self-driving Waymo cars better able to differentiate objects. You’ve also likely helped improve Google Maps by training the system to decode street view addresses.

With AI unsurprisingly figuring out how to outsmart humans in this domain too, newer versions like noCaptcha and reCaptcha v3 have already been implemented; these are more powerful and less immediately visible to the user. Able to discern whether you are a bot simply by tracking your behaviour, including frequency of clicks and typing speed, these versions only demand that you complete a picture puzzle if something in your actions is perceived as suspicious. 

US and China spearheading the AI revolution

Ultimately, it was not only Yahoo that benefitted from von Ahn’s groundbreaking cyber-security technology, with the fruits of his labour continuing to crop up everywhere – whether you’re online shopping, setting up an appointment or verifying an email address, you are very likely to come across an automated verification test. Captcha is just one of many mega-successful, internationally adopted AI projects that were launched in the US. 

In 2022 alone, image generators DALL-E and Midjourney took the virtual world by storm, and the hype around ChatGPT, dubbed the ‘best artificial intelligence chatbot ever released to the general public’ by the New York Times, seems ever-growing. It’s equally impossible to ignore the developments of the other notable AI superpower, China, which has held the biggest global share of research papers in the field of AI for some years now. The country’s enormous market size has powered some impressive leaps in the past few decades, with Baidu’s recent evolution in autonomous driving marking one recent triumph. Why is it, then, that Europe so often seems to fade into the background of the two-horse race between China and the US? And why does this matter?

The EU: slow but steady steps in AI progress

Despite Europe’s leading research community in the field and the EU’s growing yearly investments in AI, big data, and robotics, the continent has struggled to keep up with the frontrunners. Researchers have proposed a few critical reasons, including fragmentation within the EU’s digital market, inadequate commercial competitiveness, and trouble in attracting external investment. In comparison to the US and China, Europe underinvests in tech research and development, and struggles to apply their research on actual business models.  In general, Europe is restrained by a lack of strategic focus, and, with the ‘AI cold war’ narrative garnering popularity, the pressure for these weak spots to be addressed is stronger than ever. Nations that fail to prioritise AI development are without a doubt diminishing their geopolitical influence.

It’s not all doom and gloom, though. A new Joint Research Centre report reveals that the EU should be proud of its strength in AI service activities, including digital infrastructure, software, and autonomous robots. Promisingly, ensuring that Europe is ‘fit for the digital age’ was listed as one of the top two priorities of von der Leyen’s Commission, with plans to invest one billion euros a year in AI and bring overall EU spending up to 20 billion euros annually. The Council of the EU’s greenlighting of its General Approach on the Artificial Intelligence Act, the first-ever legal framework on AI addressing the ethical dimensions and risks associated with it, marks another European step in the right direction. While the act’s loopholes and inflexibility leave room for improvement, the comprehensive regulatory scheme could be groundbreaking on a global level in its safety management of the digital world, while positioning Europe as the leader in regulation of AI use.

Member states, big and small, carve their own paths

It isn’t just the EU that is implementing changes; initiatives by member states are also in the works to bridge this existing gap. The German AI Association has just announced plans to build a powerful European AI infrastructure, claiming that a ‘high-performance supercomputing infrastructure’ is what Europe needs to stay relevant and competitive. The project, called LEAM (Large European AI Model) will see ‘at least 350 million euros’ spent on building a new supercomputer. LEAM has great potential to reduce Europe’s dependency on foreign AI models, especially given that over 70 percent of foundation models are American. Germany is one of the countries at the forefront of AI in Europe, boasting many creative ideas, including the creation of Europe’s very own alternative to Captcha – Friendly Captcha. 

But even smaller countries like Lithuania are proving to be fast learners, with an exponentially growing AI community. With around 100 startups developing AI solutions, the country has performed well in international competitions like Kaggle and Apple Design Awards. In Lithuania, artificial intelligence has been used in diagnosing diseases and selecting treatment methods for many years now, with Acrux Cyber Service’s monitoring platform notably providing assistance in treating patients at Vilnius University Hospital. 

Prospects for a flourishing future

The artificial intelligence horizon in Europe, then, is one characterised by zest and flexibility. With the right discussions, out-of-the-box approaches and, crucially, an acceleration in pace, our continent might become an AI prodigy after all. Let’s hope that soon, dear reader, your engagement in the digital sphere will not just be helping Google, but also the growth and enhancement of Europe’s own AI ecosystem. Who knows? Maybe it will be you designing the next ingenious project that will put Captcha to shame.