Asher's Blog

Mahatma Gandhi appears to have seen today from yesterday.

When “science without humanity” was published as one of the seven social sins in Young India in 1925, artificial intelligence did not exist, nuclear weapons had not been used, and computers were not making decisions about employment, credit, medical care, policing or warfare. Yet the warning describes our present moment with remarkable precision.

We are now discussing artificial general intelligence, autonomous agents and the possibility of systems whose capabilities may exceed human performance across many fields. We talk about guardrails, alignment, model safety and the conditions under which a machine should be permitted to act. These are technical questions, but beneath them is a much older moral question: what happens when human beings increase their power faster than they increase their humanity?

Science expands what is possible. Humanity must help us decide what should be done, who should benefit, which risks are unacceptable and what responsibility we carry towards people with the least power to protect themselves.

Without that humanity, progress can become a more sophisticated way to cause harm.

Science is not the enemy

The warning is not against science, just as “commerce without morality” is not an argument against business. Scientific knowledge has extended life, reduced suffering, improved agriculture, connected continents and given ordinary people capabilities that earlier generations could not have imagined. Vaccines, antibiotics, electricity, sanitation, telecommunications, modern transportation and computing have transformed what it means to live.

Africa needs much more science, not less. We need better medical research, more reliable energy, productive agriculture, stronger infrastructure, advanced manufacturing, climate adaptation and technology developed for the realities of our communities. A continent with a young population cannot afford to become a passive consumer of discoveries made elsewhere.

However, a discovery does not carry its own moral instructions. Science may tell us how to edit a gene, predict a person’s behaviour, build an autonomous weapon or replace a category of human work, but it cannot by itself settle whether we should do so, under what conditions or at whose expense.

Capability answers, “Can we?” Humanity forces us to ask, “Should we, for whom, and with what protection?”

That second set of questions cannot be left until the technology has already reorganised society. Once a scientific system is embedded in hospitals, schools, banks, military operations, government services and workplaces, reversing its harms may be far more difficult than preventing them. Ethical reflection is therefore not something to add after innovation succeeds. It belongs in the design process from the beginning.

Hiroshima and Nagasaki revealed the distance between power and wisdom

The atomic bombings of Hiroshima and Nagasaki remain among the clearest examples of science advancing faster than humanity’s ability to govern its use. On 6 August 1945, an atomic bomb was detonated over Hiroshima, followed three days later by another over Nagasaki. Japan announced its surrender on 15 August, and the Second World War formally ended in September.

Historians continue to debate the necessity of the bombings, the influence of the Soviet Union’s entry into the war and the alternative paths through which Japan might have surrendered. That debate should not be reduced to a convenient sentence. What is beyond dispute is the scale of civilian suffering. Estimates commonly cited by Japanese authorities place the number of people dead by the end of 1945 at roughly 140,000 in Hiroshima and 70,000 to 74,000 in Nagasaki, with survivors experiencing injuries, displacement, illness and stigma for decades afterwards.

The bomb represented extraordinary scientific achievement. Human beings had learned to release energy from the structure of the atom, bringing together theoretical physics, engineering, industrial organisation and enormous state resources. Yet when that achievement was converted into a weapon over populated cities, the same intelligence that revealed something profound about nature also revealed something terrifying about human power.

The moral issue is not solved merely by saying that the technology worked. A weapon can perform exactly as designed while the human consequence remains unconscionable.

The survivors, known as hibakusha, have spent decades ensuring that the world does not remember the bomb only through military strategy, geopolitical calculation or scientific history. Their testimony restores the human being to the centre of the story: the child searching for a parent, the family erased in a moment, the injured person without medical care and the survivor whose suffering continued long after political leaders declared the war over.

Science without humanity converts people into targets, statistics and acceptable losses. Humanity insists that every number once had a name.

Modern weapons can increase distance from consequence

The same danger appears in contemporary conflict. Modern militaries can combine satellites, sensors, artificial intelligence, drones, precision-guided weapons and vast quantities of data to identify and strike targets from great distances. These technologies may improve accuracy and, when governed properly, may reduce some risks to civilians compared with less discriminating methods. Yet increased precision does not automatically produce humane warfare.

A technically precise weapon can still be used against the wrong target, based on flawed intelligence, at a time when civilians are present or in pursuit of an objective whose human cost has not been taken seriously. The language of accuracy can even create moral comfort, allowing decision-makers and the public to imagine that advanced warfare is clean because the machine is sophisticated.

This becomes even more serious with autonomous weapons. The International Committee of the Red Cross has argued for limits on systems that select and apply force without human intervention, emphasising the need to preserve human control and judgement. At stake is not only whether the weapon can distinguish an object accurately, but whether a machine should be allowed to determine the conditions under which human life is taken.

Human judgement is imperfect, and history gives us no reason to romanticise it. Nevertheless, moral responsibility cannot be delegated to an algorithm. If a system causes unlawful harm, somebody must remain answerable for choosing, deploying and governing that system. “The model decided” cannot become the technological version of “I was only following orders.”

The leaders developing advanced weapons must therefore consider people outside the countries financing the research. The civilian in an African, Middle Eastern, Asian or European conflict zone has the same human value as the engineer who designed the system and the official who authorised its use. National interest may explain why governments invest in military capability, but it does not erase universal human dignity.

AI makes Gandhi’s warning immediate

Artificial intelligence makes “science without humanity” particularly urgent because it is not confined to laboratories or battlefields. AI is entering ordinary decisions at extraordinary speed. It can influence who receives a job interview, whether a borrower appears creditworthy, what a child learns, which patient receives attention, what information citizens encounter and how employers measure performance.

The benefits may be enormous. AI can help doctors detect disease, allow small businesses to automate expensive processes, translate information across languages, personalise education, improve public services and give individuals access to forms of expertise that were once available only to large institutions.

However, a system does not have to become superintelligent before it harms vulnerable people. A recruitment model trained on biased historical decisions may reproduce discrimination at scale. A credit model may exclude people whose economic lives are poorly represented in formal data. A healthcare tool may perform less reliably for populations that were absent from its training set. An automated government process may deny a benefit to someone who has neither the knowledge nor the money to appeal.

The problem with automated harm is that it can look neutral. A person affected by the decision may be told that the system applied the same process to everybody, even though the process was built from unequal data and designed around assumptions that do not fit everybody’s life.

This is why the most important question is not simply whether an AI system performs well on average. We must ask how it performs for the person most likely to be overlooked.

Design from the position of the most vulnerable person

If we want to build good intelligent systems, we should begin with the most vulnerable person in society. That person may be poor, disabled, elderly, displaced, undereducated, living in a rural area, speaking a low-resource language or working in an occupation that automation could rapidly change. They may not have reliable internet, a modern device, formal identification, a lawyer, a public voice or the confidence to challenge a machine.

What happens when the system makes a mistake about that person?

A wealthy person may recover from an incorrect credit decision by using another bank, calling an influential contact or providing additional documentation. A low-income trader may lose the capital required to keep a family business alive. An executive incorrectly filtered out of a recruitment process may approach the employer through a network; a young applicant without connections may never know that an algorithm rejected them. A digitally confident customer may correct a data error, while an elderly person may simply lose access to an essential service.

The same error produces different levels of harm because people do not possess equal capacity to recover.

Humanity requires designers to account for that difference. High-impact systems should be tested across the populations they will serve, not merely on the data that is easiest to obtain. People should know when an automated system materially influences a decision, understand the basis well enough to challenge it and have access to a real appeal process. Human oversight should not be ceremonial; the human reviewer must have the authority, information and time to change the outcome.

The UNESCO Recommendation on the Ethics of Artificial Intelligence places human rights, dignity, fairness, transparency and human oversight at the centre of AI governance. The important point is not that every ethical question can be solved by a global document. It is that technical excellence alone is no longer an adequate definition of a good AI system.

A model that is impressive in a demonstration but systematically excludes the vulnerable is not advanced enough.

Africa must not inherit other people’s assumptions

Africa occupies a particularly important position in the future of AI. We have large young populations, rapidly expanding digital economies, hundreds of languages and social realities that are often poorly represented in global datasets. If most advanced AI systems are built elsewhere, the risk is not merely that Africans will consume foreign technology. The deeper risk is that we will inherit systems carrying assumptions about identity, employment, language, credit, health and social behaviour that were never designed around us.

An AI system can appear globally intelligent while remaining locally ignorant.

If a language is not represented, the people who speak it may receive a lower-quality service. If informal work is treated as an absence of economic activity, millions of productive Africans may appear financially invisible. If an identity-verification system is trained under different conditions, legitimate customers may be rejected. If automated education assumes constant broadband and individual devices, it may widen the gap between well-resourced schools and everyone else.

Africa must therefore participate in building the systems, standards and datasets that will shape its future. The African Union’s Continental AI strategy has recognised the need for an Africa-centric, development-focused approach to AI, and that direction matters. Our objective should not be to slow innovation until every possible concern disappears, because excessive caution can also preserve poverty and deny people useful technology. The objective is to ensure that innovation responds to African problems and respects African lives.

That requires investment in research, computing infrastructure, education, local-language data, safety evaluation and independent institutions capable of challenging both governments and powerful technology companies. It also requires Africans to participate in global AI governance, because decisions made in a few wealthy countries may affect workers, markets and public institutions far beyond their borders.

People who did not build the system should not be left to absorb its risks without representation.

Guardrails must survive commercial and political pressure

It is now common for technology companies to speak about guardrails, but a guardrail that disappears when it slows growth is not a guardrail. Safety becomes real when it constrains behaviour under pressure.

For a company, that may mean delaying a launch because the product can be misused, limiting a profitable capability, funding independent testing, reporting serious incidents and giving safety leaders enough authority to challenge commercial teams. For governments, it may mean preventing mass surveillance, requiring due process before automated decisions deprive people of rights, protecting citizens’ data and refusing to deploy systems whose errors cannot be meaningfully appealed.

The builders of powerful AI systems should be accountable not only for intended use but also for reasonably foreseeable misuse. No company can prevent every harmful act, and innovation would become impossible if builders were treated as infinitely responsible for everything users do. Yet it is equally irresponsible to release a capability, understand the obvious risks, profit from widespread adoption and then describe every consequence as somebody else’s problem.

Guardrails should include technical controls, but they also require governance: clear responsibility, external scrutiny, documented risk decisions, protection for internal dissent and consequences when leaders conceal harms. Ethics cannot depend entirely on the goodness of a founder or the promises of a chief executive. Systems this powerful require institutions that remain effective after leadership changes.

Policy can be necessary and still be delivered humanely

Science without humanity is not limited to scientific inventions. It also appears when leaders use data, economics and policy analysis to justify decisions without showing basic kindness towards the people who will bear the cost.

I once supported a political leader and later told friends that I could no longer offer the same support. My concern was not merely whether a particular policy could be defended intellectually. It was the manner in which policies were delivered. A policy may be economically necessary and still affect some people very severely. Leadership requires the courage to make difficult decisions, but it also requires enough decency to recognise the pain those decisions create.

Humanity does not mean that every reform must be abandoned because somebody will dislike it. It means leaders should communicate honestly, implement change carefully, provide reasonable transition periods, protect those with the least ability to adjust and remain open to evidence that the burden is greater than expected.

If a government removes a subsidy, changes an immigration rule, automates a public service or closes an unproductive industry, the policy document may describe aggregate benefits. Real people, however, experience the transition individually. One loses a job, another can no longer afford transport, a family becomes separated and a small business loses the margin that kept it alive.

The humanity of policy is visible in what leaders do for the people represented by the downside of the model.

Competence without kindness can become cruelty with excellent presentation.

“The poor you will always have” is not permission to ignore them

Jesus said, “The poor you will always have with you.” That statement can be misunderstood as an acceptance that poverty is inevitable and therefore requires no urgent response. I understand it differently. If vulnerable people will remain present in every generation, then responsibility towards them will also remain present in every generation.

Their existence is not an inconvenience in the story of progress. They are not second-class participants in humanity, and their inability to purchase the newest technology does not make their interests less important. Those of us who have education, capital, networks and influence are not more human; we are more privileged, and privilege creates an obligation to consider people who may never enter the rooms where the future is being designed.

Being our brother’s keeper means more than offering help after harm occurs. It means designing systems that do not predictably sacrifice the weak for the convenience of the strong. It means asking who is missing from the data, who cannot appeal, who loses work, who carries environmental costs, who becomes easier to surveil and who will be blamed when the technology fails.

Humanity changes the order of innovation. Instead of building first and compensating victims later, we think seriously about foreseeable harm before deployment. Instead of assuming benefits will eventually trickle down, we design access, education and transition into the plan. Instead of describing vulnerable people only as beneficiaries, we involve them in defining the problem and evaluating the solution.

Nothing about us without us is not merely a slogan for advocacy. It is a sound design principle.

Why are we building this?

Every scientific project should eventually confront a question that cannot be answered by science alone: why are we doing this?

We will not remain here forever. The engineers, investors, presidents, researchers and founders directing today’s systems will eventually leave their positions and, ultimately, this life. The technology may remain, shaping people they will never meet. If our only answer is that we wanted to prove something could be done, become the first to do it, defeat a rival or capture a market, then our ambition is too small for the power we are creating.

We innovate because human beings face problems worth solving. We want people to live longer, work more productively, learn more easily, travel more safely, communicate more freely and spend less of their lives trapped by avoidable hardship. Progress is valuable because people are valuable.

Once the person disappears from the purpose, scientific advancement becomes directionless power.

The leaders at the frontiers of AI, biotechnology, energy, defence and public policy must therefore cultivate more than technical brilliance. They need humility about what cannot be predicted, courage to slow down when the risks are unacceptable, empathy for people unlike themselves and accountability strong enough to survive commercial and political pressure.

Science should enlarge human possibility without shrinking human worth.

That is the standard. Not whether the machine is the smartest ever built, the weapon is the most precise, the policy is the most efficient or the company reaches the highest valuation. We must ask whether the advancement serves life, protects dignity, distributes opportunity and gives particular attention to the person who could be harmed without ever being heard.

Gandhi’s warning has reached the age for which it seemed to have been written. Science is becoming more powerful than anything he could have witnessed, and the distance between invention and global consequence is becoming shorter. Our humanity must now grow at the same speed.

If it does not, we may build systems capable of answering almost every question while remaining unable to answer the most important one: what became of the human being?