AI reputation is becoming a big deal in the news, but it’s not well understood. Let’s just say here it comes – The lifetime of slander you’ve always wanted.
There’s a quite funny but seriously alarming article by a guy called Kevin Roose in The New York Times about how he got a bad reputation with chatbots.
The article, “How Do You Change A Chatbot’s Mind?” is well worth reading. It’s also disturbingly interesting for another reason. Having got a negative reputation, Roose then tries to repair the damage. It ain’t easy. It took quite a bit of effort to tone down the negativity.
Nice folk tale, you might think. The “Roose Incident” got quite a lot of publicity. It might well have been the breeze before the hurricane.
There’s another much bigger and far less obvious problem. If someone uses AI to check you out, they’ll do the usual thing. They search online with the AI.
That means any negatives about you will be clearly visible to the AI. The information may be entirely false or more likely misleading. That doesn’t really matter. The AI is searching for you, and that’s what it finds.
The AI cannot check whether the information is true or false. It could try to find contradictory information, but what are the chances? You could be accused of anything, particularly by bots, and the information is weighted accordingly.
Landlords, creditors, employers, and other ornaments to human life are also stuck with this information. They may or may not believe it, but it’s now documented. Thanks to the mindless procedural environment of most businesses, it’s likely to wind up on your record.
Is this an invasion of privacy? Not necessarily. This isn’t just speculation. This information is available right now. Anyone can be searched.
This type of information comes from publicly available sources, at least at the moment. Later, it may be available for access if granted to the AI by the sources.
You can see where this is going. Financial records might be accessible to third parties. Past issues with credit, landlord disputes, workplace disputes, you name it; the possibilities are horrifying.
The possibilities are also dangerous. What if the AI finds information for people who are genuinely hostile to you? This is Doxing by other means.
How do you tell a Large Language Model that the world is full of nutters who might kill the people it’s searching for? You can do protocols, but are the protocols safe? Can they be?
How’s the AI supposed to know that something as simple as an address or phone number can get the subject of the search killed? How about blackmail on sensitive subjects?
Cheerful subject, isn’t it?
Just one example – Think how many businesses and their accountants could be in big trouble. A forensic AI search would be a true ordeal by fire. Now add the rest of the world, that is, every single person, to the mix.
There has to be a way of managing the problems somewhere in this rather unsanitary-looking Irish stew of issues. For journalists in a hyper-polarized environment, with even the word “news” becoming a four-letter word, reputation is a major issue.
For individuals, these reputation searches could be catastrophic. No credit. No mortgage. No way of answering back, either. AI is likely to be the baseline information. Do you want your business dealings or any other part of your life to be subject to this sort of scrutiny?
Regulation? Maybe, but the damage might have already been done. Civil law? Again, that’s after the event. Prevention may well be the only cure.
Look after your AI reputation, because it can destroy you.
___________________________________________________________
Disclaimer
The opinions expressed in this Op-Ed are those of the author. They do not purport to reflect the opinions or views of the Digital Journal or its members.
