The Wave Is Already Here: An Open Letter to the Lovers and Haters of AI

To everyone breathlessly telling us that artificial intelligence is going to change everything, and to everyone replying “AI slop” before reaching for the unsubscribe button:

You are both looking at the same wave.

And far too many of you are arguing about whether you like the look of it.

Artificial intelligence is behaving less like another consumer technology and more like an incoming tsunami. It is large, disruptive, difficult to predict precisely, and propelled by forces considerably bigger than individual preference. Yet our collective response has been remarkably similar to standing on the beach and shouting at the sea.

One group cheers.

Another boos.

Neither response constitutes a flood-defence strategy.

The enthusiasts tell us AI is inevitable, transformative and capable of revolutionising everything from medicine to education. The critics point, quite reasonably, towards hallucinations, plagiarism, environmental costs, displaced workers, synthetic misinformation and an internet increasingly carpeted in what has acquired the wonderfully contemptuous name “AI slop.”

Then comes the proposed solution: don't use it.

Don't buy AI-generated books. Don't read AI-written articles. Unsubscribe from companies using synthetic images. Reject AI search results. Ban students from ChatGPT. Shame companies replacing artists. Put “NO AI” on the door and hope the tide respects signage.

There is nothing wrong with any individual making those choices. Consumer resistance can influence how companies behave, and there are plenty of applications of AI that deserve criticism.

But refusing to swim does not stop a tsunami.

The uncomfortable reality is that AI no longer requires our collective enthusiasm. If an insurance company can automate a process that previously required 200 employees, it has an enormous financial incentive to investigate doing so. If a government department can process millions of documents using AI, somebody will propose it. If a pharmaceutical company can accelerate drug discovery, it will experiment. If a student can turn three hours of work into thirty minutes, some students will do it.

And if one country decides not to develop increasingly powerful AI systems, another one will.

We have crossed the point where “Should AI exist?” is the useful question.

It exists.

The question now is what we are going to allow it to do.

That distinction matters because inevitability is too often confused with helplessness.

We cannot stop earthquakes either. We nevertheless have building regulations.

We cannot abolish fire. We have fire codes.

We did not respond to the invention of cars by concluding that collisions were inevitable and therefore pedestrians should simply take their chances. We created licences, speed limits, roadworthiness tests, insurance requirements, pedestrian crossings, crash standards and rules determining who is responsible when something goes wrong.

Control does not require elimination.

AI should be approached in much the same way.

Instead of endlessly asking whether AI-generated content is good or bad, we should be establishing provenance standards so that people can know how something was created.

Instead of universities desperately trying to detect ChatGPT from a student's writing style, we should redesign assessment around the capabilities we actually want students to demonstrate.

Instead of waiting until synthetic research contaminates scientific literature, journals should establish rigorous disclosure, verification and data-provenance requirements now.

Instead of debating whether a convincing AI-generated video is technically a “deepfake,” lawmakers should be deciding what happens when synthetic media is used to impersonate someone, manufacture evidence or manipulate an election.

Instead of artists having to discover that their work has disappeared into an enormous training dataset, we need meaningful rules governing consent, attribution and compensation.

And perhaps most importantly, we need to decide what remains human.

Not because humans are magically incapable of producing rubbish. Several thousand years of recorded history should probably disabuse us of that idea.

But because some activities have value beyond the efficiency of their output.

A child struggling to write an essay is not an inefficient essay-producing machine. The struggle is part of the education.

A therapist listening to a frightened person is not merely an unusually expensive natural-language interface.

A researcher wrestling with contradictory evidence is doing something different from requesting a plausible synthesis of it.

If we do not identify these distinctions now, economics will eventually make them for us.

That is the danger I think both AI evangelists and AI opponents underestimate.

The greatest threat isn't necessarily that AI becomes extraordinarily intelligent. It is that AI becomes extraordinarily cheap.

Cheap enough that replacing the human becomes the default decision and retaining one requires justification.

Once that happens, “AI slop” stops being an amusing insult and becomes an inadequate description of a structural problem.

There may even come a point when objects created before this transition acquire a peculiar significance. A textbook published in 2017, a photograph taken in 2004, an academic paper written in 1996 or a newspaper archive from 1982 possesses something we may increasingly struggle to establish about future material.

We know the technological environment in which it was created.

Not necessarily that it is true. Not necessarily that it is good.

But that it came from the other side of the wave.

That should bother us.

Because we still have time to ensure that future generations do not need a publication date before 2022 simply to have confidence about where information came from.

The answer is not to smash the machines. Nor is it to hand them the keys and congratulate ourselves on embracing the future.

It is provenance. Disclosure. Auditing. Liability. Regulation. Education. Authentication. Professional standards. Independent oversight. And, where necessary, prohibition of particular uses.

Some applications should be encouraged.

Some should be regulated.

Some probably should be illegal.

And those decisions should be made deliberately rather than emerging accidentally from whatever happens to be profitable.

So this is an appeal to both sides.

To the people who love AI: inevitability is not an argument against regulation.

To the people who hate AI: rejection is not a substitute for regulation.

You do not have to like the wave.

You do not have to celebrate it.

You do not even have to use it.

But please stop pretending that shouting at it will make it turn around.

The wave is already here.

We cannot stop the water.

We can still decide where to build the walls.