Do we control AI or does it control us?

The answer to this basic question that has been worrying regulators, consumers, industries is neither of the two options. There is a third element in the chain of control and that is the owners, developers or investors in the companies that have invented and marketed AI.

They are in control and these people are few in number and are very powerful whether they are in Silicon Valley or in China.

The recent events involving GROK, specifically its ability to digitally undress individuals or depict them nude or in altered attire without their consent, merely at a user’s request, is a case in point. In fact, a number of countries have decided to ban GROK from their markets.

GROK is owned and run by Elon Musk, who also owns a large number of very important satellites in the management of surveillance and data-gathering as well as space rockets. Musk is probably one of the most dangerous persons on earth and there are a few others like him.

These persons must come under the thumb of the regulators before they get out of control.

In fact, GROK has been banned in two or three countries because of undressing and posting naked pictures of children or women or men without their consent or knowledge, since this constitutes a crime and an insult to morals in many countries. What did Musk do? He did not remove the possibility of his AI tool undressing or redressing living persons against their will. He only blocked this function in the three countries that had banned it.

What does this choice of action by Musk tell us?

It tells us that AI does not work on its own accord or according to the whims of the consumer who stupidly decides to use it for these purposes. It depends on the algorithms that manage it. The algorithms are created and written by people upon instructions of the owners of the tool.

All AI as well as all other apps and digital tools are blind and stupid. They only obey the algorithms that their creators decide to put in. These can be changed, taken away or blocked and the tool will be unable to perform what it did before the changes.

A few years ago, I suggested that the EU, as the regulatory authority in our part of the world, introduce legislation requiring the establishment of an algorithm bank, into which copies of all digital products or services would be placed. The EU already has a digital police unit to watch for malicious apps but they have no access to the algorithms themselves that command and direct these malicious, illegal or spying products and services.

All AI as well as all other apps and digital tools are blind and stupid

By forcing all existing or new products to lodge a copy of all the algorithms in their products in this bank, it would allow the EU specialists to check the suspicious products and services before they cause damage and not react after the fact, after our children are forced to suicide, or our criminals to blackmail people with false photos, or foreign powers to spy on us via their car digital cameras.

Prevention is better than cure.

Such a bank would also allow copyrighted works of art of music or literature to be protected against copying without payment because such copying tools would be banned outright or forced to alter their algorithms and the owners fined.

Mustafa Suleyman, the CEO of AI at Microsoft, the company for which I served as vice president for legal and government affairs for some years, was very blunt when he warned all companies developing more and functionalities of AI to ensure that they do not do enough to create control functions of the tool while developing more and more functions that can get out of control.

He believes that since “you cannot steer something you can’t control”, we must not only learn to limit AI but also to make sure that AI cares enough not to harm humans. According to him, and he knows far more about what is happening through the said algorithms behind the scenes than any of us, scientists are confusing the job of containment of AI with the job of aligning AI. There are two different philosophical challenges that we must get right.

One is about limiting AI and enforcing limits that will keep AI within pre-determined boundaries. The other addresses whether AI will act in humanity’s best interests or not.

First comes containment and he is right.

The EU should take the initiative in the world to either work with those scientists such as Suleyman or, even without such cooperation, just go ahead and force all products using AI or even just using the internet to lodge their algorithms for control by the EU experts.

Those who fail to comply should be barred from entering the EU single market or, if already operating within it, required to exit.

We need to defend ourselves from the risks of nefarious AI and of owners only interested in making money or, worse still, who are carrying out a planned policy to control humanity rather than improve it.

With such controls in place, much like those governing pharmaceuticals, Europeans can reap the many benefits these tools offer to humanity while preventing criminal misuse and irresponsible developments.

John Vassallo is a former ambassador to the EU.