As artificial intelligence slowly takes over many regulatory processes – starting perhaps with registration, identity, and qualification checks, then moving on to analysis of complaints and discipline data – we should see great improvements in efficiency and possibly cost reduction.

At the same time, AI will also be moving further into health care than it has already done: reading digital scans more accurately than humans ever can, replacing surgeons with computer-controlled robots, and eliminating face-to-face consultations for diagnosis. In engineering, architecture, and legal services, AI is already playing significant (if mainly analytical) roles.

It is not fanciful to imagine a future regulatory complaint about an error by a robot or software program, but who will be accountable if AI causes harm? The person who wrote the software, or the hardware that it ran on, or the organization which used it, or the human who switched it on? I can only imagine the complexity of determining responsibility in such a situation. One thing we can be sure of is that AI will not be responsible because AI is not an entity – at least not yet.

Despite the misleading attempts by those who promote AI assistants to us – by giving it a name like Alexa, Siri, or Claude and by using ”I” when it answers, as though it was a person – AI is the product of a machine, not a person. Despite a name, it has no self. Up to now we have been able to hold responsible the humans in charge of a machine for a failure that causes harm, but if the machine runs itself, acts by itself, and thinks for itself, who is the responsible person?

As AI moves deeper into more and more areas of activity, displacing humans as it goes, it will no doubt eliminate most errors, but what will happen to judgment, empathy, and compassion, those human variables which are interpretative, intuitive, and instinctive rather than rational?

There is much discussion at the moment about “compassionate regulation,” or as Zubin Austin simply puts it, ”kindness.” The contrast here is between artificial intelligence and natural intelligence. The latter includes emotion, insight, and imagination. Without those elements AI will never replicate the way we humans make our judgments and, no doubt, our mistakes.

AI of course does learn, that’s why it is so powerful. Could it learn to replicate feelings? It won’t experience those feelings, but it may learn to take into its algorithms the kinds of feelings that a human decision-maker might have. We know that AI is already capable of something that looks like imagination because it is able to invent fake references for articles it writes for students cheating on their exams. I don’t think it is yet able to express compassion toward the person harmed as well as the person who caused the harm, as a human decision-maker can.

I think humans will be needed for a long time to do the things we humans do: act irrationally, emotionally, with kindness towards others and, sometimes consciously, make mistakes because of a greater good.

There is nothing new about my musings here on this relationship between human and machine. Over 200 years ago, in 1818, Mary Shelley published “Frankenstein.” In her novel, Victor Frankenstein builds a creature in his laboratory based on a new and previously unknown science. The monster he creates is huge, powerful, and (unlike AI) has emotions. The monster attempts to fit into human society but is shunned, which leads it to seek revenge against its creator. In another contrast with AI, the monster has no name. Shelley denies it human identity.

We have humanized AI with a name but no emotions. That contrasts with Shelley’s monster, which had emotions but no name. In her novel it is emotion – the monster’s desire to relate to humans, to be valued – and its rejection by humans which causes the monster to turn against its creator and to destroy him.

I wonder if the real step forward, or maybe backwards, will be if the developers of AI not only create machines under human control that can destroy humans, as military drones are doing in Ukraine and Gaza right now, but create machines that can express pleasure or horror at what they are doing.

If AI ever has feelings, maybe it will be able to practice compassionate regulation on our behalf, but maybe it will prefer to make its own choices. It was the very fact that Frankenstein’s monster had feelings that ultimately made it uncontrollable.

That then, was fiction. This now, is not.

Harry Cayton is a sought-after global authority on regulatory practices who created the Professional Standards Authority (PSA) and pioneered right-touch regulation. He is a regular Ascend Voices contributor.