A Conversation for Humanity — Article Three | 24 October 2026
By Deverout Graham
Managing Partner, Deverout & Associates
Bedford · London · Global Affiliates
The Intelligence We Create, and the Humanity It Reflects
What We Teach, We Sometimes Meet Again
My fellow human beings, there is a curious phenomenon in our relationship with the world. We teach a bird to imitate our voice, and one day we hear our own words returned to us. We may raise or train a lion until it recognises a particular human being, yet familiarity does not make it a domestic animal or remove the dangers of its natural behaviour. We shape what we encounter, and then we are surprised when something of our influence appears before us.
I call this the Paradox of Being Human: we create systems to extend our abilities, then encounter our own ambitions, contradictions and limitations in what those systems produce.
Artificial intelligence gives this paradox a new significance. We train models using enormous collections of language, images, code and other information produced or selected by people. We refine some systems through human judgments and feedback. Then we ask the machine to speak, reason, advise and create. Often, something recognisably human comes back.
A Reflection Is Not a Human Being
The analogy has limits, and those limits matter. A bird is a living animal. A lion has biological drives, an evolutionary history and the capacity to act independently of its trainer. An AI model is not a bird or a lion. It does not acquire an animal’s instincts by learning from data, and fluent conversation does not establish that it possesses human experience, intentions or moral understanding.
Nor is an AI system a perfect mirror of humanity. What it produces depends on the information it encounters, how that information is filtered, its training objectives, the judgments used to refine it, and the instructions and tools involved when it is deployed. A mirror may reflect, but the builders choose its shape and where it points.
Research from Stanford on whose opinions language models reflect found that training data and human feedback both shape model behaviour; additional tuning did not necessarily make models more representative of public opinion. That is an important warning against assuming that an AI answer speaks for all humanity. Read the Stanford policy brief (20 September 2023).
What Does the Machine Learn About Us?
Consider what the human record contains: scientific discoveries and superstition; kindness and cruelty; cooperation and conquest; beautiful literature and deliberate misinformation. AI systems may reproduce, transform or amplify patterns found in those records. But this is not destiny. Responsible choices about data, evaluation, design and oversight can alter outcomes.
Here is the uncomfortable question: if a machine repeats a prejudice found in human history, whose responsibility is it to challenge that prejudice? The machine’s output is the immediate problem. The larger responsibility belongs to the people and institutions that build, procure, deploy and govern it.
The US National Institute of Standards and Technology identifies risks including fabricated outputs, harmful bias, privacy concerns and misplaced human reliance in its Generative AI Profile (26 July 2024). These are practical governance issues, not merely philosophical worries.
An Objection: A Mirror Can Distort—and Discover
I must challenge my own metaphor. If we call AI a reflection of humanity, we risk understating its capacity to produce useful combinations and insights that no single person supplied. AI can help researchers explore possibilities, support translation across languages and reveal patterns in large collections of information. It can be more than a passive repetition of its sources.
But originality of output is not proof of truth. A plausible explanation may be wrong. A confident recommendation may omit a crucial human circumstance. Knowledge requires verification, not admiration of eloquence.
There is another objection. Perhaps a system trained on humanity’s contradictions cannot escape them. I reject that conclusion as too fatalistic. Human beings also inherit flawed ideas, yet education, debate, evidence and institutional reform can change what we do. Our technologies should be designed to support the same discipline of correction.
The Business Transformation Test
For business leaders, this philosophy has immediate consequences. Human-centred AI governance must begin before deployment, not after harm occurs. Four questions should guide any serious AI transformation strategy:
- Whose information and experience are represented—and whose are missing?
- How will we test for errors, bias, privacy risks and unequal outcomes?
- Who has authority to question, override or stop an AI-supported decision?
- What evidence would cause us to revise or withdraw the system?
These questions convert philosophical humility into operational responsibility. They also reinforce a first-principles approach to business transformation: define the human purpose first, then select and govern the technology accordingly.
The Courage to Recognise Ourselves
The most revealing discovery of the AI era may not be that machines can imitate aspects of human expression. It may be that humanity is forced to examine the assumptions it has placed into its institutions, information and technologies.
If we want AI to support cooperation, we must reward cooperation in the systems around it. If we want fairer outcomes, we must test for unfairness. If we want knowledge, we must welcome evidence that corrects us. And if we want technology to serve life, we must make that responsibility visible in our decisions.
Winning is good. Participation is everything. Participation, however, must include the courage to listen when the answer is inconvenient.
We talk. We listen. We look. We learn.
A Question for Humanity
When an AI system reflects a harmful human prejudice, should our first response be to correct the machine, change the institution using it, or confront the human assumptions behind both? I invite disagreement, evidence and thoughtful contributions. That is how conversation becomes knowledge.
Deverout Graham
Managing Partner, Deverout & Associates
Transformation from First Principles
Bedford · London · Global Affiliates
deuerout.com
Leave a Reply