14 August 2025

Welcome to the machine: Toll fiasco another example of tech tyranny

| By Ian Bushnell
Join the conversation
20
Sydney roads

Ka-ching: faceless technology drives today’s toll roads and other businesses. Photo: Sardaka.

Turns out AI isn’t that smart after all.

Pinging Canberra drivers whose vehicles were nowhere near Sydney toll roads is pretty dumb after all.

The number may have been the same – thanks, NSW, for forgetting about us Yogis and duplicating the Y series – but the scanners could not even pick up the clearly marked ACT on the plates.

Then again, AI can only be as clever as its training, so maybe that’s on the humans.

READ ALSO Sydney toll giant ‘sincerely apologises’ for billing Canberra drivers who never used roads

Trouble is, this sort of consumer nightmare has been bubbling away for years and will only get worse as automation accelerates and becomes further entrenched in how we not just do business but live our lives day to day.

The interface between consumer and provider has become increasingly anonymised as governments and businesses, always seeking cost savings and efficiency gains, rely more on technology, cutting humans out of the equation.

Our business relationships have become perfunctory as the automatic payment systems run like clockwork, scooping up billions of dollars with relentless efficiency.

Taxes and charges are raised, landing in our inboxes, often without raising an eyebrow and, if not already sated by direct debit, paid with a few careless clicks.

For the privilege of spurning cash, we are hit with a card surcharge that most of us wear because of the convenience and its relative insignificance.

But multiply all those millions of transactions and it is BIG money.

No one can doubt that technology has changed our lives for the better, making commerce easier and saving us time in travel and queues.

The downside, though, is that dealing with the relentless logic of machines, even the increasingly sophisticated chatbots that flash up from the screen, is not the same as dealing with real people. However, some call centre staff are not much better, especially if they work off tightly written scripts or are based overseas.

On the other side of the counter, the machines coldly calculate with efficiency but little understanding.

For many people, fatigue sets in, and a deferential indifference takes hold, unquestioningly accepting a charge or a payment.

READ ALSO A road-user charge for electric vehicles is just another cash grab

So the great online cash register clicks on.

And, as the Sydney toll issue shows, when a customer notices something does not add up they face the Kafkaesque ordeal of (1) trying to contact somebody, (2) trying to convince a low-level functionary that their tech has made a mistake and (3) trying to recover their losses even when it is clear the fault is theirs.

As AI advances, it is hard to believe that these sorts of situations will improve. Is consumer law keeping up? What kind of regulation can protect consumers? What obligations should companies have? Can they be enforced?

It is hard enough to avoid spammers and scammers without having to be ever alert to the voracious appetites of tech-driven companies that rely on set-and-forget attitudes and business models on autopilot.

Our policymakers and legislators need to get on top of this and stay ahead of the curve, not be left in technology’s wake playing catch-up.

Free Daily Digest

Want the best Canberra news delivered daily? We package the most-read Canberra stories and send them to your inbox. Sign-up now for trusted local news that will never be behind a paywall.
Loading
By submitting your email address you are agreeing to Region Group's terms and conditions and privacy policy.

Join the conversation

20
All Comments
  • All Comments
  • Website Comments
LatestOldest

Let’s use a Greek letter, or invert the Y.

Tom Worthington4:49 pm 15 Aug 25

Reading number plates is the sort of project a second year computer student could do at university (I know as I teach them). The problem of duplicate number plates could be solved with a very simple software upgrade, which checked across motor vehicle registries for duplicates, and flagged them for manual processing. Also high resolution digital cameras are now cheap, so the system could be upgraded to read what state is on the plate.

This duplicate problem only occurs where the vehicle doesn’t have an electronic tag (they have unique identifiers) and the number plate has to be read optically. The motorist is charged extra for this “video matching”. However, clearly there is no human intervention, so what is the fee for? The extra cost of the digital optical processing would be thousandths of a cent. Perhaps motorists should initiate a class action to recover all such fees charged, which are thousands of times the cost.

ANPR (Automated Number Plate Recognition) is not (repeat: not) AI. It is a well studied process to distil symbols from an image; an algorithm. The ability to search a database, given the output from ANPR is not AI either.

Not The Mama2:23 pm 15 Aug 25

The error here seems to be pretty dumb – If the cameras can read the numbers in the number plate, they can also read where the plate is from.

Even so “AI” has a lot of improving to do to earn its name. I’ve spent some frustrating hours with Microsoft’s Co-Pilot and with Antropics Claude. In my case I thought these AI offerings could save me hours of C++ coding effort. I have concluded that if these two popular offerings are anything to go by, then the day that AI takes over the world is still a far far far….. far into the future – Both AI produced code, and it worked, but when something needed tweaking then they would both forget the original task, write additional code that was not compatible with the original code and so on. On one occasion I asked for a critical review and corrections on a lab report that I wrote. and the AI re-write the method section, saying that I had used equipment and methods that I had never ever used (clearly the AI copied the method from someone else in the net).

As far as I am concerned – the performance does not merit the hype. They are cool tools as long as the user is aware of the limitations. I also think that companies that are currently firing staff and replacing them with AI may find themselves hiring again before too long.

Looks like a private sector version of Robodebt.

It seems that it’s the people, not the machines. Those in charge usually overstate what the programming does e.g the algorithm might just calculate 1+1 but is spun to be the “new future of automation”. One thing those in charge hate is being found out, as Robodebt showed.

How would toll road users feel about queuing up to manually pay at each toll booth? Solves the problem of incorrect registration plates being charged. Imagine how much time that would waste… then we could complain about that too!

True, but I think Ian’s point is that consumer law needs to keep up, with more responsibility on the automators. One can waste quite a bit of time chasing a small individual debt yet if ignored those small errors add a lot of pure profit for the organisation.

Ian, AI called, said they are hiding in your house and they are going to get you.

@nobody
Highly likely that AI dialled the wrong number, nobody 😊

I suspect companies and organisations running these types of devices are happy to push the risk and cost of monitoring over to the consumer. Just like they’re happy to have useless, frustrating support chatbots so the company can claim to provide help to their customers.

Daily Digest

Want the best Canberra news delivered daily? Every day we package the most popular Region Canberra stories and send them straight to your inbox. Sign-up now for trusted local news that will never be behind a paywall.

By submitting your email address you are agreeing to Region Group's terms and conditions and privacy policy.