Silicon Valley wants you to pick a side: will artificial intelligence deliver total utopian salvation or trigger a sci-fi apocalypse? In the latest edition of Delve into AI, ADONIJAH NDEGE cuts through the binary hysteria. From a book launch in rural Kenya to the high-stakes boardrooms of Anthropic, Ndege explores why both AI evangelists and doomsayers are missing the unglamorous, everyday realities of human existence. An algorithm might discover a life-saving medical breakthrough in a California lab, but can it fix the supply chains, clinical shortages, or sky-high costs that keep that cure out of reach for a rural clinic?
Artificial intelligence (AI) could make meaningful improvements to human life. The possibility that it might help doctors detect diseases earlier, accelerate scientific research, or make knowledge more accessible deserves to be taken seriously.
These would be substantial achievements, particularly in places where access to expertise is expensive or scarce.
Yet, I am increasingly uncomfortable with the gap between those modest possibilities and the grand promises made for the technology. Improving specific aspects of daily existence is one thing; suggesting that intelligent machines will produce a superior human existence requires a far more demanding argument.
My concern is that we are mistaking advances in raw capability for progress towards outcomes we have barely defined. What does it mean for AI to improve human life? Whose life improves? And how do we evaluate whether the benefits justify the costs?
These questions lingered during the September 12 launch of A Theology of Suffering: Quest for Meaning at the Ukarimu Retreat Centre Chapel in Molo, roughly 200km west of Nairobi, Kenya. The author, Fr Melchior Marandu, a Tanzanian Jesuit priest, acknowledged technological advances, poked holes into their power to alleviate human suffering.
“What is the meaning and cause of suffering and evil, death, that technological and scientific advances cannot eliminate? Even AI cannot eliminate death or suffering,” he said.
“AI is growing and growing; more than what we are using. We are using very little,” he added, arguing that its expanding capabilities could not resolve the fundamental questions raised by suffering.
I would be cautious about treating this as a definitive indictment of AI’s technical utility. No technology has had to eliminate suffering entirely to justify its existence. Medical advances that grant people more time with their families are profoundly valuable; an AI system that helps discover such treatments deserves recognition.
Nevertheless, Marandu’s remarks expose a critical question often drowned out by technical hubris: how much of the human condition can technological progress actually repair?
A hard question to ignore
That question becomes harder to ignore when set against the warnings emerging from within the industry. The expectations on AI now stretch from extraordinary improvements in human life to the possibility of human destruction.
Jacob Coxon, a former employee at Anthropic and OpenAI—the two leading AI developers—warned on September 8 that AI could potentially kill humans.
“The people building AI earnestly believe that it could kill us all by the end of the decade,” Coxon warned, after publishing his resignation on X.
It’s hard to establish the likelihood of such an outcome from that warning. We should be careful not to present an expression of belief as a reliable forecast.
Still, a claim of this magnitude from an industry insider demands scrutiny, especially when the public is also being asked to embrace the technology’s potential benefits.
In what could be seen as a response to Coxon, Dario Amodei, the co-founder and chief executive of Anthropic, said in an essay titled We Must Pace the Frontier remains optimistic about AI’s potential to improve medicine and prosperity, but argues that safeguards need time to catch up with capabilities.
“We must slow the pace at which we improve the capabilities of AI models,” he wrote.
His account has a personal experience. His father died of a disease that became curable a few years later; Amodei himself survived an early-stage cancer. These experiences possibly explain the urgency he attaches to medical progress.
He also acknowledges that “we still only understand a tiny fraction of what goes on inside these models.” He proposes external evaluation and coordination among companies and governments. His argument is that the benefits remain worth pursuing, but that the speed of development must allow for better safeguards.
I see three questions emerging from these arguments from Marandu, Coxon, and Amodei. They touch on what AI can achieve for humans, whether we can manage its consequences, and how its achievements would translate into better lives. Each needs an answer of its own.
Salvation or catastrophe
The first question is about capability. Can a system perform a task reliably enough to be useful? This is where claims should be most concrete. If an AI tool is intended to help diagnose an illness, we should be able to examine its accuracy, the circumstances in which it fails, and whether using it improves care.
The second is about control. A system may perform a task well while creating risks through its operation or use. Better performance does not settle whether it is safe to deploy. The more consequential the application, the stronger the case for independent scrutiny and clear responsibility.
The third is about human welfare. Even a capable, adequately regulated system may deliver uneven benefits. Its existence does not ensure that the people who need it can afford it, that institutions can use it effectively, or that the savings it creates will reach the public.
I think this third question receives too little attention when the discussion moves towards either salvation or catastrophe.
Consider what it would take for an AI-assisted medical breakthrough to improve the life of a patient in a rural community. The discovery matters, but so do approval, production, distribution, clinical staff, and the patient’s ability to pay.
Failure at any of these stages can prevent a scientific achievement from becoming a personal benefit.
I do not need to share every part of Marandu’s theological argument, but I understand his wariness. A service can attract customers and generate revenue while leaving unresolved questions about its wider effects.
Equally, a useful technology need not transform civilisation to be worth developing.
The promise of AI must become more specific as its capabilities grow. We require clear empirical evidence of benefit, independent audits of risk, and public institutions equipped to ensure that advancements reach beyond those best positioned to pay for them. Above all, we must retain the right to challenge specific applications of technology without being forced into a binary choice between blind endorsement and total rejection.
Between salvation and catastrophe lies the everyday existence most people are trying to improve. Most young people simply seek decent work, accessible healthcare, and basic security. If AI can meaningfully improve those everyday realities, that will be an achievement truly worth defending.
True scale demands moving beyond surface-level integrations to robust execution. We’ve filtered the noise out of Moonshot 2026, optimising the conference strictly for high-calibre connections between startup founders, global financial operators, enterprise leaders and individuals rewiring Africa’s technical frameworks. Get 20% off Early Bird tickets for a limited time.
