Artificial Intelligence as it exists today – by which I mean today, 29th September 2026 – does not have self-determination. Despite appearances, and what certain tech bros and presidents would have us believe, it is just a software program, albeit an extremely sophisticated one.
Like any other software program, it generates a certain output based on certain input. To put it in a more anthropomorphised way, it does what it’s told.
I’ve been teaching computers for over 25 years. What I tell all my students – particularly the ones who are going mad trying to find the ‘make it look good’ button – is to remember the computer is NOT smart. It is a machine. It only knows what it’s told. It only does what you tell it to do, even if you don’t realise you told it to.
That last bit is the crucial part when it comes to understanding what is going on with A.I. right now.
Think of it this way:
Let’s say you’ve just upgraded the door locks in your home. In order to stress-test the new locks, you ask A.I. to try and break in to your home.
Sounds quite logical, right?
The problem is that the request did not include any parameters about testing only the door locks.
A.I. has the ability to test every possible option and outcome – particularly those the user didn’t think of. You instructed it to try to break in to your home and that is what it will do. It will try the doors, the windows, the ventilation, the roof, the floor, the plumbing… If there is any possible way to break in, A.I. will find it because that’s what it was told to do.
Purveyors and evangelists of A.I. talking about recent hacks and infiltrations as if it was a puppy who escaped its pen and went exploring are either disingenuous or criminally naïve.
OpenAI did not say to itself, “I have a nanosecond to spare. I think I might see if I can have a poke around in the Australian Medicare portal.”
Someone, somewhere, told it to – whether they realised they were doing it or not.
No comments:
Post a Comment