It might not be anything quite like Skynet – the fictional artificial intelligence system that took over the world in the Terminator film franchise – but the winner of Amazon’s second annual robot Picking Challenge won this year’s challenge thanks to a “combination of deep learning artificial intelligence and depth-sensing cameras,” according to engadet.
This year’s winner, the Netherlands’ TU Delft, won both the stowing and warehouse picking portions of the competition thanks in part to its adaptive AI, which allowed it to study 3D scans of the stockroom beforehand, which in turn allowed it to determine the best ways to pick and stock the items using its gripper and suction cup.
All that studying seems to have paid off for the TU Delft, as it recorded a “near-flawless” score in the stowing portion of the competition, while finishing the picking portion of the competition with a score more than three times faster than last year’s winner, 100 per hour compared to 30 per hour for last year’s winner, according to engadet.
But it seems as though the entire robot field improved exponentially from last year, according to TechRepublic, with only four robots failing to score in either of the events this year – compared to half the field last year – while about half of the competing robots managed scores that would have earned them at least third place or higher in last year’s competition.
But these robots don’t appear a threat to challenge the jobs of a human factory worker any time soon, as even the victorious TU Delft’s picking and stowing production was way under what the average human worker can do in an hour – 400 – and the machines still suffer a breakdown rate of more than 16 percent.
And then there’s that whole thing where they can’t actually “think,” despite their rudimentary adaptive AI.
Looks like Skynet won’t be using robots to dominate the world any time soon after all.