The IT industry loves its acronyms. Sometimes to the point of creating ones that no-one will ever remember. The CAPTCHA acronym is short for “Completely Automated Public Turing test to tell Computers and Humans Apart”. Very catchy.
The purpose of CAPTCHAs is to prevent bot attacks and spam. They do this by creating a challenge. If you pass the challenge, you are human.
Can you read this?
The text-based CAPTCHA displays a distorted image of randomly generated characters. You must type the characters you see in the text box.
I had to pass the CAPTCHA test on the CIPC site this week. I failed a few times. Is that the number 1, or a lower-case letter L? Is it a zero or the letter O? Is it a lower-case Q or a lower-case G? Am I human?
For more than 10 years, bots have solved these CAPTCHAs faster and more accurately than us. So I don’t know what the CIPC site confirmed, other than its abililty to frustrate me.
How much of a bicycle?
Then there is the image-based CAPTCHA.
“Select all squares with bicycles”. This is clearly a trick question. How much of a bicycle? This can lead to esoteric questions of existence. Is it a bicycle if it is only the tip of a handlebar?
I hate these. Most people do. People have described them as frustrating and depressing. One writer wrote that they sucked the life force from his body. That might be an exaggeration, but not by much.
They may soon be a thing of the past. Last year researchers from ETH Zurich announced they had built an AI model that can solve these image puzzles 100% of the time.
Are you human?
Then we have the checkbox that asks you to confirm that you are not a robot.
It’s a bit dehumanising to have to confirm that you are human. It’s even more depressing when we understand why this works. It’s because we are less efficient than bots at checking the box.
Invisible CAPTCHAs work in a similar way. They analyze user behaviour, including browser history (yikes!) to decide if you’re real.
How long will this work? The researchers at ETH Zurich created a special mouse movement model to imitate human movement. They used fake browser and cookie information to make the AI agent appear more human.
The next step
The next step is AI-powered verification. Yes: we will use AI to prove that we are not AI.
This will apparently lead to a new business opportunity: Human Verification as a Service (HVaaS). Google’s already doing this with its I-am-human checkbox.
Remember those sci-fi books and movies where no-one knows if you are a human or a robot? We’ve reached the digital equivalent.
I’d love to hear your comments.