logo
Vincent Ramdhanie
Senior Software Engineer
Back to Articles

The Promise of Artificial Intelligence

beginner
By Vincent Ramdhanie•
ai technology society ethics
The Promise of Artificial Intelligence

The experts have sorted themselves into camps: some say AI will end us, some say it will save us. I think both camps are looking at the wrong thing. The technology is a tool. The dangers are real, but they mostly come from us — and that is precisely why they can be addressed.

It is hard to follow the news about artificial intelligence without noticing that the people who build it cannot agree on what it means. Nobel laureates warn of extinction. Chief executives promise abundance. Both groups include serious people with deep knowledge of the field, which makes the disagreement itself worth examining before taking a side.

The doom camp

Geoffrey Hinton, who received the Nobel Prize for the neural network research that made modern AI possible, puts the chance of AI leading to human extinction within thirty years at 10 to 20 percent. His reasoning is simple: we have never had to manage something more intelligent than ourselves, and there are few examples of a more intelligent thing being controlled by a less intelligent one.

Yoshua Bengio and colleagues, writing in Science, argue that we are not prepared for the risks of increasingly autonomous systems and that safety research commands a small fraction of the attention that capability research does. Eliezer Yudkowsky and Nate Soares take the maximalist position — their book is titled If Anyone Builds It, Everyone Dies — and call for development to stop entirely. And in 2023, hundreds of researchers and most of the leading lab executives signed a one-sentence statement declaring that the risk of extinction from AI should be a global priority alongside pandemics and nuclear war.

Nor is the concern confined to elder statesmen looking back at the field. The week I drafted this article, a researcher resigned publicly from Anthropic, urging the industry to slow down before AI learns to improve itself and writing that "the people building AI earnestly believe that it could kill us all by the end of the decade." Whatever one makes of the claim, the fear is live inside the labs, not just outside them.

The euphoria camp

The other camp is just as credentialled. Sam Altman describes a "gentle singularity" in which superintelligence arrives almost quietly and delivers scientific progress at a pace we can barely imagine. Marc Andreessen argues, without much hedging, that AI will save the world, and that slowing it down is the real moral failure. Demis Hassabis speaks of "radical abundance" — the end of disease and of scarcity itself. Ray Kurzweil has kept the same schedule for decades: merger of human and machine intelligence by mid-century.

Interestingly, some of the most prominent voices stand in both camps at once. Dario Amodei of Anthropic has written perhaps the most detailed optimistic case — Machines of Loving Grace, which imagines a century of medical progress compressed into a decade — while running a company organised around the premise that the risks are serious. That double position is itself informative: the people closest to the technology mostly describe a fork in the road, not a destination.

The middle

A third group holds that both extremes mistake a tool for a destiny. Andrew Ng has long compared AI to electricity — transformative precisely because it is general and unglamorous. Arvind Narayanan and Sayash Kapoor argue that AI should be understood as normal technology: powerful, but diffusing through institutions at the speed institutions actually move. Daron Acemoglu's economic modelling reaches a similar temperament — meaningful gains, but distributed according to choices we make, not laws of nature.

This is the camp I find most convincing, and I want to explain why in plainer terms than the economists use.

Every powerful technology has carried its dangers

The benefits of electricity are not in question today, yet electricity can and does kill people. The benefits of the automobile are not in question, yet well over a million people die on the world's roads every year. We could extend the list through aviation, medicine, chemistry — every technology that has advanced human welfare has also brought genuine harm.

I do not say this to be dismissive about risk. People dying on roads or from electrical faults is not acceptable, and we should be doing everything in our power to eliminate those harms. But notice what we actually did in each case. We did not call for the end of electricity; we invented circuit breakers, insulation standards, and licensing for electricians. We did not abolish the car; we engineered seat belts, crumple zones, traffic codes, and driving tests — and deaths per mile fell by an order of magnitude. The pattern is consistent: where a risk can be predicted, it can be mitigated, and the work of mitigation has always cost far less than the benefits the technology delivered.

AI belongs in this pattern. The failure modes being documented today — systems that deceive, that are misused for fraud, that embed bias, that consume enormous energy — are predictable, and predictable risks are engineering problems and policy problems. That includes the environmental cost: the International Energy Agency projects data centre electricity demand roughly doubling by 2030, and that cost deserves the same treatment as tailpipe emissions did — measurement, standards, and relentless efficiency work, not resignation and not denial.

None of this happens on its own. Safety engineering is a deliberate investment, and the doom camp is right about one thing: the investment is currently small relative to the stakes. Where I part company with them is the conclusion. The response to an under-engineered technology is more engineering, not abandonment.

The danger that worries me most

If the technical risks are tractable, one risk is not primarily technical at all: concentration. A tool this general, held by a few corporations or a few governments, is an instrument of oppression waiting for a user. History does not lack examples of what follows when transformative capability pools in few hands — and this particular capability touches speech, work, surveillance, and warfare all at once.

The Bahá'í International Community put the underlying principle well in a statement to the United Nations on digital technologies: "Like any tool, it can be deployed productively or destructively, depending on the ethical considerations underlying its design and use." And the remedy follows from the same statement: "Every member of the global community should benefit from the fruits of the human mind, be it equitable access to technological innovation or the knowledge generated through its creation." (Reflections of Our Values, 2021)

Equal access is not charity; it is the safety mechanism against the misuse I fear most. All nations and peoples need access to this capability — private corporations and governments alike, rich countries and poor ones alike. A world where intelligence-as-a-tool is broadly held is a world where no single holder can use it to dominate the rest. The engineering analogy holds here too: distribution is to concentration what the circuit breaker is to the electrical fire.

A tool in service of one humanity

Why do I believe the benefits are worth all this work? Because of what the tool actually amplifies. Bahá'u'lláh wrote: "Regard man as a mine rich in gems of inestimable value. Education can, alone, cause it to reveal its treasures, and enable mankind to benefit therefrom." (Gleanings) A technology that can tutor any child, translate any language, and accelerate any science is, at its best, an instrument for exactly that revealing — at a scale no previous generation could attempt.

The same writings carry the caution that belongs beside the promise: "If carried to excess, civilization will prove as prolific a source of evil as it had been of goodness when kept within the restraints of moderation." (Gleanings) That sentence is over a century old and reads like it was written for this moment. The technology is not the moral agent in the story. We are.

Which is why the deciding factor will not be model capability. 'Abdu'l-Bahá described religion and science as "the two wings upon which man's intelligence can soar into the heights" — the point being that technical power without moral development flies in circles, at best. Call that second wing what you like: ethics, governance, spiritual maturity. Without it, the pessimists' scenarios become plausible. With it, the optimists' scenarios become achievable — and shared.

"The earth is but one country, and mankind its citizens," Bahá'u'lláh wrote. (Gleanings) That is the standard I would hold this technology to. Not whether it is impressive, but whether it helps us act like citizens of one country — and that outcome was never going to be decided by the technology. It will be decided, as it always has been, by us.

References

These thoughts are my own. I had help with grammar, structure, and research from Claude.