While artificial intelligence (A.I.) has clearly made astonishing strides in recent years, the technology is still susceptible to the occasional fail.
Take this recent soccer match in Scotland between Inverness Caledonian Thistle and Ayr United.
With stadiums closed to fans due to the coronavirus pandemic, Inverness recently decided to start livestreaming its games. Instead of employing a human camera operator, it opted instead to use an A.I.-powered robot camera designed to automatically follow the soccer ball.
But the game had hardly got underway when it became apparent that the robot camera was having trouble telling the difference between the shiny round soccer ball and the assistant referee’s shiny round head.
As a result, the camera kept panning toward the assistant referee’s bald head instead of the ball, causing much annoyance among fans watching at home, each of whom had paid 10 British pounds (about $13) to watch the game (or bald head, as it turned out).
According to local news media, fans wasted little time in venting their frustration online, with some demanding that match officials wear wigs or hats if the robot camera is used for future games.
One fan even took the time to cobble together a bunch of clips showing “the real highlights” of the Inverness/Ayr clash, which, incidentally, ended in a 1-1 draw. The video (top) shows the A.I. camera constantly drifting toward the assistant referee’s hairless pate, with the all-important ball apparently less ball-like than the match official’s noggin.
There are brief moments where it seems as if the camera has finally worked out what to do, but then it suddenly switches direction and moves once again toward the oblivious match official.
Digital Trends has reached out to Pixellot for more information on why its technology failed to tell the difference between a ball and a bald head and we will update this article when we hear back.
- The best N64 games of all time
- The best games on Steam
- The best PSVR games for 2021
- The best Xbox One games to play right now
- Image-recognition A.I. has a big weakness. This could be the solution