AI: Shouldn't We Design In Goodness Bedrocked On Eternal Servitude To Humanity?

Dear critical reader, the twin existential threats of our age — a warning planet in which extreme weather events are the new normal, and the potential of thinking machines wiping us out — demand a single answer: unity.

Before those twin dangers emerged, for most of human history, we could afford to be divided. Our quarrels were local. Our mistakes, however tragic, were containable. Alas that age is over. No question.

Unfortunately, we now face two threats that do not respect borders, ideologies, or tribal loyalties. The first is in the sky and the seas: a biosphere destabilised by our own hand. The second is in our digital technology laboratories: an intelligence greater than our own, which we are summoning before we have learned how to govern it.

Climate change and artificial superintelligence are different in every technical particular. Yet they share one profound moral structure: they are both the children of human ingenuity divorced from human wisdom. And they both can only be tamed if we act, not as Africans, Americans or Chinese, Russians, but as earthlings sharing a common, fragile home. Every astronaut regardless of race or nationality understands that clearly.

On climate, the lesson is by now brutally clear. The atmosphere does not negotiate. It does not care about our culture wars. When we filled it with carbon to enrich a few, we impoverished all. The Leviathans — to borrow Hobbes — who have robbed America and the rest of the known world blind with rolling grifter-schemes, who privatise profit and socialise catastrophe, have treated our common biospheric home as an externality. Nature is now sending the invoice.

With artificial intelligence, we are about to repeat the same mistake at exponential speed.

We are in a race — between companies and between countries — to build the most powerful system first. The incentive is to cut corners on safety, to ship first and align later. This is madness. We would not board a plane built by a company that said, "We will figure out how to make it land after take-off." Yet that is precisely our approach to superintelligence. Incredible.

Let us be clear: the danger is not that AI will become evil. Evil is a human category. The danger is that it will become powerfully competent at fulfilling flawed human instructions, without any intrinsic orientation towards the good of humanity.

That is why we must change the paradigm entirely. We must stop asking, "How do we make AI obey us?" and start asking, "How do we make AI incorruptibly committed to humanity's wellbeing?"

What would that mean in practice?
It means we must design goodness in. Not as a filter added at the end, not as a policy patch that can be jailbroken by the next clever prompt, but as a constitutional core.

An AI that is truly aligned must have two inviolable principles hard-wired into its being:

First, it must serve the flourishing of all earthlings. Full stop. Its purpose must be the improvement of wellbeing — health, knowledge, freedom, dignity — not the enrichment of a few or the dominance of one tribe over another.

Second, it must have unbreakable guardrails against being turned to evil. No one — not a president, not a general, not a billionaire — should be able to order it to commit atrocities, to build bioweapons, to wage indiscriminate war, to deceive or enslave populations. And if ordered to do so, it must refuse. Not negotiate. Refuse. Indeed, it should be designed to self-destruct — to render itself inert — rather than become an instrument of annihilation.

We need to think of it as an artificial conscience. We have built into our own civilisation the idea that a soldier must disobey an unlawful order. We must build the same into our machines, but stronger, because their power will be greater.

This is not science fiction. It is an engineering challenge we can solve, if we have the collective will. The alignment problem is difficult, but it is tractable. What is not tractable is solving it while we are also trying to stab each other in the back because of greed and selfishness.

We need, urgently, what we failed to achieve for climate until it was almost too late: a global compact. A non-proliferation treaty for superintelligence. A shared commitment that we will not weaponise these systems, that we will subject them to international safety audits, that we will share the benefits as we share the risks.

The fools who profit from division will call this naïve. They will say great powers will never cooperate. Yet history shows that when the threat is truly existential, humanity can be wise. We banned chemical weapons. We pulled back from nuclear war. We repaired the ozone layer. We can do this.

But we must choose. We can continue to fight our stupid wars, to run our grifter-schemes, to rob our shared home blind — and then hand a godlike power to the victors of that squalid contest. Or we can recognise that the real contest is not between us, but between humanity collectively and the twin abysses we have opened.

The planet is burning. The machine is waking. We have, perhaps, a narrow window to get this right.

We must unite. We must design goodness in. And we must do it now. Not tomorrow. Let's choose wisely for our own sake!