Why License Plate Recognition is not A.I.
I got a fine yesterday. No, sorry, not a parking fine, it's a bill from Wilson Parking Enforcement Services (the "Border Force" for shopping car parks). The bill is for liquidated damages. In other words, the shopping centre could have had another car, people and fresh buyers using the same bay.Oh it was all very clever, really. But it's not A.I. Wilson Parking has automated its Parking Enforcement Service. Without people being involved, I was auto-fined. To be fair, a person did leave the infringement notice under my wiper. I'm sure one day that will be a long-armed robot. Today? A human with a thankless job. A perfect fit for A.I.
Let machines do all the work, so that people have time to think.[IBM's Pollyanna Principle (1977)]
Let's get one thing straight
License plate recognition technology is not A.I. License Plate recognition (Optical Character Recognition) has been around since 1978 - we've had it for 40 years. It started about five years before IBM's Personal Computer hit the shelves. This was about the time when the Atari 400 (for children) was released.Oh "Rogue" - the hours I spent roaming your stony rooms.Anyhoo... At the Carine Glades Shopping Centre - where I'm posted today (I'm getting a bit leery about this face-to-face selling thing but I must persevere for a bit) there are parking cameras all around the place. Saves inspectors time if an auto-detect camera can see you parking, take a pic and then dob you in to an inspector via SMS. It's not A.I. because essentially no actual decisions are being made. A.I. is concerned with intelligent computer programming and decision-making algorithms. So anyway, I got busted. I had no idea you had to pay for parking and saw no signs - but I got done anyway. And if this article about parking appeals and Wilson Parking is anything to go by (courtesy my mate Rob), I've got Buckley's chance of keeping my $65. I've spoken to Centre Management and we shall see. While this isn't an example of A.I. in action, it soon will be. And it will be a negative experience because of the algorithms companies like Wilson will have written.
The Wilson Parking Enforcement Service Algorithm
- Take photo when a car parks
- After 4 hours check to see if car is there
- If there, send parking inspector robot and place ticket under wiper
- Allow 30 days for appeal
- increase fine after 28 days
- replay to appeal after 28 days when fine increases by $20
- charge larger fine
- If fine not paid take legal action
- If car gone in less than 4hrs, do nothing
Good AI / Bad AI
But all this argy-bargy got me thinking. How is AI perceived? The whole point of AI is to mimic (and then ultimately replace) people, so there's already a leery public out there. A public who are big Terminator film fans. Elon Musk thinks AI is humanity's biggest existential threat. Stephen Hawking is not a big fan, either - but he also believes intelligent aliens will invariably be malevolent (he supports listening out for signs of life over transmitting our location). Bill Gates now thinks AI will be really helpful to humans but he wasn't so optimistic in 2015. One way to sell AI to people is to personify it. To make it sound "cuddly". AI is like a child, eager to learn, keen to play and experiment.Uncanny Valley
And then there's uncanny valley. Anything trying to resemble a human, but doesn't quite get there (like Sophia in the video below) illicits a creepy feel.Here's another great example of an A.I. being sold as a human baby. It doesn't work. It's not right.
Comments