I was asked to select the boxes with traffic lights to prove I wasn’t a bot.
I’ve done it hundreds if not thousands of times, but this
time I thought: "If AI is so smart, why can't it solve a simple CAPTCHA?"
We writers are a curious lot, and a minimal amount of
research unveiled that the traffic lights aren't really the test.
It seems that modern CAPTCHAs don't care much whether you can spot a
traffic light. They care whether you behave like a human while doing it. How
you move the mouse. How long you hesitate. Whether you click with just enough
uncertainty. It's less an eye exam than a vibe check.
So, while AI can often recognize the images with no problem, it
isn't built to impersonate a person navigating a website. The pictures are
mostly stage props.
The CAPTCHA is like a nightclub bouncer who isn't
checking your ID anymore. He's watching how naturally you reach for your
wallet.
The fact that AI struggles with a CAPTCHA, doesn't tell me
much about AI. It does, however, indicate we're redefining intelligence: more than Can you
solve the problem? but also Can you look enough like your average human
while solving it?
Unquestionably, AI will move past this and eventually we'll have a new hoop to jump through to say, "I'm human. No really, I am."
_________________________
CAPTCHA rejected this suggested update:


No comments:
Post a Comment