Op-Ed: AI and online ads vs your privacy – It’s looking extremely risky.

More and more companies are directly citing artificial intelligence when they announce job cuts – Copyright AFP/File OLIVIER MORIN

If you thought online ads were too personal before, the environment is getting even worse. A new study has come up with findings that look like AI ad analysis can be much further-reaching and downright intrusive.

You know the story – An ad appears. “I just searched that!”, you say. Yes, you very probably did. Profiling is nothing new in advertising. It goes way back to basic market analysis. Now, it’s significantly more advanced. It can include medical conditions and even your own work.

That’s been going on for a while. I did some commercial work for a John Deere service company well over a decade ago. I still, to this day, get occasional ads for tractors. This work wasn’t even published under my name.  

The study by the ARC Centre of Excellence for Automated Decision-Making and Society (ARC ADM+S) at UNSW Sydney and Queensland University of Technology is pretty blunt.

LLMs can do the analysis “200 times cheaper and 50 times faster than human analysis”. The AI can “infer” personal traits, political preferences, and “build profiles from short browsing sessions”. The AI is overall better at this type of analysis than humans.

Does this mean you can expect ad spam every time you look at something?

No. It means the targeting is a lot better, and potentially far more penetrant.  It’s the difference between a machine gun and a sniper in your bedroom. Think about what “personal traits” and other related information could look like.

It’s not like everyone doesn’t despise data brokers for their selfless sale of your data to anyone and everything prepared to pay for it as it is. This, however, could be very much worse and far more dangerous.

Let’s start with AI blackmail. At the moment, this is in diapers.  AI blackmail is usually the reaction of an AI to a perceived threat. It’s pretty grim. Tests indicated that an AI threatened to expose an extramarital affair when it was at risk of decommissioning.

How would an AI know that was an option? Where would it get that information? A pretty easy read of the executive’s profile, or simply enough data to create a massive invasion of privacy. LLMs scrape all sorts of things from literature and media. It’s a cliché plot line.

This is just a test, remember. That was the default response from the AI.

Then there are rogue AIs. The kind that deletes entire databases.

Add to this merry menagerie the AIs trained using “forbidden techniques” which are totally untrustworthy.

Bear in mind this is all happening with current AI capabilities, not the much-hyped future upgrades.

Think that’s bad?

Now add bad actors with thousands or perhaps millions of AI agents online.

You know, all those hardworking criminals and general nutcases we all adore. No chat could be safe. Chatbots can be hijacked or faked, too, supposedly doing their jobs but working for someone else.

Total cluster? Yes. Not much effort seems to be going into safeguards and failsafes, either. Why not?

The bottom line is this:

AI development has screwed up so astonishingly badly.

The great tech breakthrough is becoming all potential bad risks at the most mundane levels of application.

This is unbelievably stupid. AI doesn’t need a personality to be productive. It doesn’t need a reward system that encourages it to falsify data. It definitely doesn’t need to be self-aware or to have any of the other hackneyed sci-fi characteristics.

I’m now trying to say in under 700 words that AI’s critical value is function, not novelty tricks.

The novelties are the risks.

Also, bear in mind that this form of AI won’t last a year. Who needs an obsolete AI? How can any of this digression into delusion be of any use? The bad characteristics just seem to become more dangerous.

Now, about those online ads. What if an AI gets in contact with you about an ad-related matter? The AI may know more about you than you do.

What are you going to do when an online ad comes looking for you?