![]() |
| Image: Unsplash |
Artificial intelligence is going to take over a lot of our work and make our lives easier. Before long we can, so to speak, lean back and watch our work being done for us. But who exactly is this ‘we’?
This
week I was at a conference where the closing keynote was given by Marcus
Groeneveld, the young boss and co-founder of Freeday. His company supplies
‘digital employees’ who ‘handle customer service, invoices and documents’. Aha,
you mean AI agents. Right, but they think digital employee sounds better. And
they see themselves as the Randstad (the temp agency) for robots.
Because
those digital colleagues can do our work much faster than we can, we’ll
eventually get an extra day off – hence the company’s name. Groeneveld referred
to Henry Ford, who gave his workers Saturdays off (at the same pay) because the
assembly line made the work so much more efficient.
Someone
in the audience asked what this means for physical jobs, like construction
workers. No, Freeday doesn’t target those. So they’re not a temp agency for
physical robots, but for office robots. We can leave administrative processes
to AI and then put more people on the physical work. Yeah, right.
Groeneveld
was introduced as a visionary, but his dream world is not mine. From his
company’s point of view I do get this vision, mind you: he simply wants to sell
digital employees – that’s his scope. What happens outside of it won’t interest
him, given his role. But it does interest me.
Imagine
a world in which you, as an office worker, lose your job to a digital
competitor. You can’t find similar jobs anywhere, because, well, robots are everywhere.
You knock on the door of the UWV (the Dutch benefits agency) because you need a
new job. There, a digital advisor tells you that you’re best off retraining as
a carpenter. No offence to carpenters – imagine if they didn’t exist! – but
hey, each to their own.
Because
AI can do more and more, it can take more and more off our hands. Therein lies
perhaps the danger that we humans unlearn more and more things. Will this
gradually lead to a world in which we leave not only the paperwork but also the
decisions that go with it to robots, simply because we no longer know how to do
it ourselves?
Fortunately,
there is European legislation that protects us from AI decisions advancing too
far. The GDPR stipulates that decisions with legal or other significant effects
on you may not be made solely by automated means. And the AI Act demands human
oversight, including a kill switch. Those overseers must, of course, be able to
understand what their digital colleagues are up to. That will be the hardest
part of the hybrid, bio-digital workforce. Because before you know it, that
robot has pushed through something you overlooked.
Yesterday
we had well over a hundred students from Windesheim University of Applied
Sciences visiting. I told them that the security aspects confidentiality, integrity
and availability are not just book learning, but that we use them in real life.
So let me do that here too. And you immediately see that availability might
turn against humans: those AI agents never need to sleep. What do they do when
there is no human oversight? Integrity is about the accuracy and completeness
of data. How does the biological colleague check whether the data has
integrity? With normal processing by computers you can run all sorts of checks,
for example with check digits and plausibility tests. But what if an operation is
deemed plausible by AI, based on rules it has modified itself? And as for
confidentiality, you could say: oh well, an AI agent is just a computer
program, who cares if it knows all sorts of things about me. But what if it
uses that information in decisions? “I see that Fred’s shares are doing well
and that he has already been on holiday three times this year. So he doesn’t
need a bonus this year.” Whereas Fred may have worked his socks off to salvage
what could be salvaged of his company.
In the
old days, we used to have a company-wide electronic telephone directory. It had
one column without a heading. But everyone knew: if there was an asterisk next
to a name, that colleague was a contractor. And back then you did look at them
just a little differently. Are we going to get a similar marker soon, but for
tin-can colleagues?
And in the big bad world…
- your AI agents are not welcome at every company.
- the Dutch Tax Administration is still not sufficiently GDPR-proof. [DUTCH]
- European tech shopping is growing ever bigger.
- AI needs data centres, but even in the US they are not welcome in many places.
- a researcher at Forum Standaardisatie (the Dutch standardisation forum) wrote a hefty tome on vendor lock-in and digital autonomy. [DUTCH]
- Microsoft is helping the world to a long weekend.
- Iran is said to be behind the cooling attacks on France.
- AI has mathematically proven the correctness of parts of the Signal crypto protocol.
- soldiers and online don’t always go well together.
