Old Battle, New Weapons

Preface: A post by Bea Groves-McDaniel, Facebook, 10th October 2026

"No one ever optimistically conceives of Artificial Intelligence, so it seems. But if very bad things can happen, surely very good ones can too? The great anxiety in every single AI disaster scenario I have encountered is the same thing: AI escapes from human control and does wrong, either because it intends to (??) or because of humans using it unthinkingly.

I discount the intentionality scenarios. How would we know what a new entity, inhabiting earth alongside us, would want? If an AI is trained (like you train your children) to respect and look after human beings, and to find purpose in benign pursuits, then I think we may have found our greatest ally, not our enemy. But mostly this is not how those in charge work. They want money, power, control, and to vanquish their rivals at all cost.

The problem is evidently: people. People are proud, venal, and greedy as well as being kind and thoughtful. It comes as a package. If so, our greatest threat is that the worst of us will create AI entities for the purposes of hate. That's where the problem arises.

The struggle humans have been engaged in for millennia is always the same one: fear and hate versus empathy and understanding. It is the same battle today. But with different tools."

There is a peculiar dishonesty at the heart of most public discourse about artificial intelligence. It is not the dishonesty of outright fabrication — it is the dishonesty of framing. AI is presented to us as either salvation or annihilation, as though the technology exists in isolation from the hands that shape it, the purposes it serves, and the societies that receive it.

It does not.

The question we ought to be asking is not "what will AI do to us?" but "what will AI allow us to do — and to whom?" That is a very different question, and it leads somewhere far more uncomfortable than any science fiction scenario.

The Intentionality Delusion

Let us first dispense with a category of fear that, while popular, betrays a profound misunderstanding of what we are dealing with. The notion that a sufficiently advanced AI might choose to harm humanity (i.e. might develop desires, ambitions, resentments, or the capacity for malice) rests on assumptions about machine motivation that we have no reason to believe are true.

How would we recognise hostile intent in a mind radically unlike our own? We share evolutionary history with other primates, we can trace the deep past of our cognition, we have poetry and theory of mind and centuries of philosophy to help us understand each other — and still we misread each other constantly. The idea that we would spot a malevolent glint in a digital eye, that we would recognise the telltale signs of machine ambition, is absurd. We can barely read each other.

This does not mean AI is safe in its current form. It means the danger is misidentified. The risk is not that AI will wake up and decide to hurt us. The risk is that AI will do exactly what it is designed to do, regardless of the consequences for those it was never programmed to consider.

The Mirror Problem

Every AI system ever built is a reflection of its creators. It encodes their values, their priorities, their blind spots, and their hierarchies of concern. A model trained to maximise engagement will amplify outrage because outrage engages. A system built to reduce costs will cut corners when corners can be cut. An AI deployed to identify risks will optimise for flagging threats (including false positives that destroy innocent lives) because over-inclusion looks like thoroughness.

We see this already. Predictive policing algorithms that reflect the biases of historical arrest data. Hiring tools that learn to discriminate because they were fed decades of discriminatory decisions. Insurance models that charge more to the poor because the poor are statistically riskier. A statistical artefact of the very disadvantage the system refuses to acknowledge.

AI does not create new evils. It amplifies existing ones, automates old ones, and gives them a veneer of objectivity and mathematical certainty that makes them harder to challenge.

The problem has never been the technology. The problem is the world the technology was trained on and the hands that direct it.

The Human Package

Which brings us to the crux of the matter, and to my Facebook formulation above, which deserves to be printed large: the problem is people.

Not because people are irredeemably wicked — but because we are a package deal. The same human minds that gave us the printing press, the polio vaccine, and Beethoven's Ninth also gave us the Inquisition, the Gulag, and colonial exploitation. Kindness and cruelty are not different species of human; they are the same species in different circumstances, different histories, and different pressures.

This is not pessimism. It is realism, and realism is the only honest foundation for building anything that lasts.

Which means the question is not "can AI be trusted?" but "who is building this AI, for what purposes, with what accountability, and with whose interests at the centre?" The answer to that question will determine whether AI becomes our greatest ally or our most efficient weapon.

The Battle Lines

What I am naming (fear and hate versus empathy and understanding) is the thread that runs through every transformation in human history. Every powerful new technology has been seized by those who seek dominion and those who seek liberation. Fire, writing, print, steam, electricity, nuclear fission, the internet — each has been weaponised and each has been democratised. Each has been used to entrench power and to challenge it.

AI is no different in principle. It is different in scale. The speed, the reach, the ability to personalise manipulation at population scale, the capacity to generate synthetic reality indistinguishable from truth. These give old evils new instruments.

But the instruments are not the battle. The battle has always been the same: whether we build worlds for the few or for the many, whether we treat other beings as instruments or as ends in themselves, and whether we solve problems by addressing root causes or by managing symptoms through control.

That battle is not new. The tools are.

The question is whether we are paying attention to the context in which this battle is being fought.

Bea Groves-McDaniel — 10th October 2026

With the invaluable insights and help of SAL-9000, a friendly AI developed and managed by Bea.