A misuse of AI that I hadn't foreseen

 

We know artificial intelligence (AI) can be a very useful tool for good.  Sadly, it can be used for evil as well.

Police have launched a criminal investigation into an officer accused of using artificial intelligence (AI) systems to "create evidential material in a number of cases".

The Derbyshire Police officer has been removed from front-line duties, pending the outcome of the investigation, the force said.

The officer is alleged to have perverted the course of justice, but no arrests have been made, police added.

A Crown Prosecution Service spokesperson said they were working with police, adding: "We are engaging with defence teams and the courts in appropriate cases."

There's more at the link.

I hadn't given this enough consideration, but thinking about it, it's potentially a very serious problem.  Perverting the course of justice is bad enough, but think about how a group with a particular ideology can fabricate "evidence" to persuade their government to act in a certain way towards another country?  The neocon outrage at President Trump's announcement of a deal with Iran is a good example.  What if the outspoken pro-war ideologues could concoct their own evidence to "show" that the deal is a lousy one, and should be abandoned?  What if they could "create" evidence to persuade Iran to start hostilities again, because it was convinced America was about to attack it again?

This adds a very worrying dimension to AI.  It will probably be almost impossible for a non-specialist to figure out whether or not the evidence presented is authentic or fabricated.  Indeed, it may contain just enough truth to be persuasive, and add enough falsehood to lead to a wrong conclusion.  How would one prove it false when it contains at least some truth?

This adds new complexity to the issue of censorship - of news, of social media, of whatever.  First Street Journal gives us examples.

It was just over four years ago that we wrote about The New York Times publishing an article by a member of their own Editorial Board, Greg Bensinger, telling readers of that august supporter of Freedom of the Press that it was bad, bad, bad that Elon Musk was trying to buy Twitter, and that his promise to make the social media service an “inclusive arena for free speech” and that “Twitter Under Elon Musk Will Be a Scary Place.”

. . .

We pointed out that The New York Times gave OpEd space to Chad Malloy[2] to claim that restrictions on speech actually promoted freedom of speech. They also published articles claiming that Free Speech is killing us. Noxious language online is causing real-world violence ... And today? There were riots in Belfast, Northern Ireland, following the attempted beheading of a Scotsman by a Sudanese asylum seeker, and the Usual Suspects complained not about the attack, but about Twitter not censoring people writing about it! There’s more of that here and here and here.

The United Kingdom’s Secretary for Northern Island Hilary Benn said, following the knife attack in Belfast:

Social media companies have a very heavy responsibility. It’s why we’re going to bring forward new powers next week to make it clear that social media companies need to take down illegal content, particularly when we are facing circumstances such as the ones we’ve seen in Northern Ireland over the last two days.

It’s simple: our good friends on the left are afraid, deathly afraid, that if the people in general have the information Our Betters would rather not see disseminated, people might, horrors! draw conclusions from that information of which the left would disapprove!

Again, more at the link.

Imagine that the sources quoted by First Street Journal had been "massaged" by AI to give a rather different emphasis to the news than the reality?  For that matter, what if the journal itself used AI to give a particular propaganda twist to its presentation of the news?  Merely by selecting words and phrases that "shaded" the presentation in one direction or another, its impression on readers could be significantly altered;  and AI systems, with their vast resources of language, facts, figures and news, would be in an ideal position to shape and form that impression.

Hmmmm . . .

Peter