Are you a robot? Amazon patent reveals it’s getting harder to tell

You know those little tests you have complete to prove to a website that you're not a robot? Welp, robots may soon be better at those tests than humans.

That’s the takeaway from a patent awarded to Amazon this week.

The tests, called CAPTCHAs, can be aggravating. Sometimes I can’t tell if a photograph includes a stoplight or not. If I were shown that photograph 10 times, I might answer differently every time. But a robot would get it right.

And remember that endless internet controversy about whether a dress was blue or white? A robot could tell you exactly what color it was.

Amazon’s patent, named "Turing test via failure," suggests a new way of thinking about those tests. Maybe in the future we won’t identify humans by their superiority over robots. We’ll identify humans by their failures and inconsistencies.

Other fun examples of human failure from the patent:

• Humans are easily tripped up by word association. We tend to incorrectly answer the question, "What do cows drink?" by answering: "Milk." Same with the question "What do you put in a toaster?" The correct answer is bread, not toast. A bot would never make those mistakes.
• In a photograph, coal may be brighter than snow, based on how the coal reflects the light. But a human will usually answer that the coal is darker, regardless of what the photo shows, because we understand that in most other circumstances, coal should appear darker than snow. A bot would simply measure the brightness of each as it appears in the photo and answer correctly.
• Humans accept "close enough" answers where a bot could not. Asked who said, "I don't think we're in Kansas anymore, Toto," a human would think of the movie The Wizard of Oz and answer: Dorothy. A bot would recognize that Dorothy never said those exact words. What Dorothy actually said was "Toto, I've a feeling we're not in Kansas anymore."