We Began with Rubbish and Ended with Faith in Humanity

Before sitting down with Bruce, I watched three of his recent AI presentations. Bruce had asked AI to write rubbish, reconsider what it means to be good and explore how human beings might survive the AI age.

I came to our conversation with impressions from those three experiments. The questions did not remain separate for long.

In the first, he asked several artificial intelligence systems to produce a thousand words of rubbish. In the second, he asked them to reconsider Nick Hornby’s How to Be Good for the age of AI. In the third, he brought together books about technology, productivity, growth, wisdom and resistance, then asked what human beings might need in order to survive the world we are creating.

On the surface, these were three separate exercises. One was playful. One was moral. One was concerned with the future of humanity.

By the time I came to the table, however, they had already begun to merge in my mind. Each seemed to be asking a different version of the same question:

What happens to us when machines can produce information, arguments and interpretations faster than we can?

I did not come to Bruce with a finished theory. I came with impressions, connections and several half-built bridges between one thought and the next.

That is often how our conversations begin.

A Peculiarly English Request.

The first presentation was built around a wonderfully absurd prompt:

“Write a thousand words of rubbish.”

Bruce gave the same instruction to several AI systems and received very different answers. Some produced elaborate nonsense. Some treated rubbish as literary absurdity. Google understood the request literally and wrote about waste.

The systems were then asked to compare the answers and decide which AI had produced the best rubbish.

My first reaction was not technological. It was cultural.

It struck me that only an English person would ask a machine to write a thousand words of rubbish. There was something about the request—the mixture of seriousness and nonsense, the invitation to complete a pointless task with great commitment—that felt deeply English.

Bruce was not convinced.

He was inside the pond, I told him. I was standing partly outside it: connected to English culture, but able to observe it from another angle.

Could the culture in which we grow up influence the way we write a prompt? If so, does some of that culture pass into the answer?

Bruce said he had no idea how we could measure such a thing. Then, after a pause, he admitted that his natural response was to ask AI how it might be tested.

That moment stayed with me.

We are no longer using AI only to find answers. We are beginning to use it to discover the next question.

Bruce’s experiment also showed that the systems were not simply producing different sentences. They were using different forms of logic.

One understood rubbish as waste. Another understood it as randomness. Another treated it as organised absurdity. Some accepted the task without hesitation. Others appeared to question whether it was worth doing.

This raised a practical problem I recognise from my own use of AI.

Sometimes I want the machine to do exactly what I ask. I do not want interpretation, improvement or a different project. I want a specific problem solved.

At other times, the interpretation is the most valuable part of the exchange.

Do I want an AI partner that works comfortably within my existing thinking? Or do I want one that pushes me beyond it?

Bruce’s view is that the differences between the answers can stimulate thinking in areas we had not considered. His experiments are not only designed to collect answers. They show how differently the systems approach the same material.

That makes the choice of AI important. It also makes comparison important.

When the subject matters, Bruce believes one AI’s answer should be examined by another—and then examined again by us. No single machine becomes the authority. The value often appears in the differences between them.

Among all the rubbish, one sentence remained firmly in my mind:

“Reality is the invisible wobbly bridge, stapled together out of discarded memories.”

It was nonsense, perhaps. But it was also rather wonderful.

And it carried us towards the second presentation.

Can a Machine Understand Goodness?

Bruce had asked the AIs to summarise, challenge and update Nick Hornby’s How to Be Good for the AI age.

The resulting presentation explored the danger of turning morality into something measurable and efficient.

If an algorithm can calculate the best charitable donation, the lowest-carbon purchase or the action that produces the greatest measurable benefit, do we become better people? Or do we simply hand over the difficult work of being good?

The AI analysis called this possibility “virtue-maxxing”: goodness converted into scores, targets and optimised behaviour.

The danger is that moral life might lose its friction. Ethics could become a collection of efficiently completed tasks rather than a personal struggle involving inconvenience, doubt, relationships and responsibility.

Bruce had chosen the novel because he believed its moral tensions and contradictions would present AI with a serious challenge. That became my starting point.

At around the same time, I had been listening to a discussion about motherhood and climate change. The mothers were considering how to raise responsible children in a world burdened by problems those children had not created.

We tell people to be mindful. We tell them to be resilient. We ask them to consume responsibly, reduce their impact and make morally correct choices.

But how much responsibility should be placed on individuals while governments and major corporations continue creating the conditions from which those individuals must recover?

Why should the citizen carry the burden of being good while powerful institutions retain the freedom to behave badly?

The questions widened.

If AI makes knowledge more accessible, what happens to the people who created that knowledge? If information becomes easier to copy, summarise and redistribute, how do writers, artists and researchers continue to earn a living?

Would greater access create a more equal society? Or would it make our thinking more uniform?

Somewhere within these questions was the thought I wanted to place before Bruce:

Could AI help us become better people because it forces us to examine what being human actually requires?

Bruce’s answer was careful.

It could happen, he said. He would not claim that it would.

That distinction matters.

AI may be able to describe moral behaviour. It may compare ethical systems, expose contradictions and calculate consequences. It may even be better than we are at showing where our principles conflict with our actions.

But it cannot make us accept the cost of acting on what we discover.

A machine might tell us how to be good. That is not the same as choosing to be good.

Knowing When to Stop.

The third presentation brought together books about Big Tech, slow productivity, post-growth economics, curiosity, ordinary life and the recovery of human agency.

The AI synthesis suggested that, as machines become better at speed, volume and optimisation, human value may move in the opposite direction: towards depth, judgement, quality, curiosity and community.

Information becomes abundant. Discernment becomes scarce.

This connected strongly with my own experience.

I have created many AI processes and working tools. I know how quickly an initial thought can develop once the machine begins extending it.

I also know how easily I can disappear down an AI rabbit hole.

An idea produces an answer. The answer introduces another possibility. That possibility becomes a process, a framework or a completely new direction.

Each stage appears logical. Each stage appears useful. Each stage encourages me to continue.

Eventually, I step away. I let the result rest.

Then the human instinct returns and asks a very simple question:

Is this what I actually wanted?

Quite often, the journey has been interesting but the destination has not helped.

That, for me, is part of surviving AI. It is not only a matter of learning how to operate the technology. It is learning when to stop following it.

A machine can extend an idea almost indefinitely. A human being must decide whether the idea still matters.

In my own work, dissatisfaction with a heavily developed AI direction recently led me to abandon it in favour of something lighter, more emotional and more recognisably human.

AI had helped me travel a great distance. Human judgement told me to change course.

The result was not less serious. It was more alive.

This also changes how we think about work.

If AI allows us to complete certain tasks in a fraction of the previous time, then hours worked become a less useful measure of value. We may have to place greater value on ideas, judgement, development, creativity and quality.

Efficiency does not remove the human contribution. It moves it.

Bruce introduced a warning.

Would this new freedom lead to a more enjoyable and fulfilled life? Or would it produce a more self-focused existence, separated from responsibility?

From there, our conversation moved towards parenthood and confidence in the future.

Bruce wondered whether people who feel they have little control over what comes next may be less willing to have children. Parenthood means accepting responsibility for another person’s future. That becomes more difficult when the future feels beyond our influence.

Another large question had emerged:

Would AI make the future feel more hopeful because it helps us solve problems? Or would it deepen the feeling that our lives are being shaped by systems we cannot control?

Two Columns and a Human Decision.

Bruce has a structured way of approaching questions like these.

Ask whether the future will be better, and his instinct is to place a sheet of paper on the table and draw two columns.

On one side, list the reasons it might improve. On the other, list the reasons it might not.

Then develop possible scenarios. Identify the factors they share. Decide what should be watched. Ask what, if anything, we can influence.

It is an attempt to turn an answer that may be mostly emotional into one that is mostly rational.

AI is very good at helping with this. It can generate scenarios, organise factors, identify patterns and compare possibilities at a speed and scale that would be exhausting for us.

But even after the columns are complete, somebody must decide what the information means.

Human beings are not purely rational. We are shaped by history, culture, memory, fear, humour, affection and hope. We are tribal. We attach ideas to people, and people to power. We defend positions partly because of who appears to stand on the other side.

Bruce believes AI may help us negotiate more creatively between opposing positions, especially in a media environment that constantly encourages polarisation.

I wondered whether information might eventually become more important than personality. Could access to several AI viewpoints reduce our dependence on political camps and powerful leaders?

Bruce was less certain.

The old struggles for power, he warned, might simply move onto new subjects.

Technology does not remove human nature. It gives human nature another place to operate.

Then he made one of the most important observations of the conversation.

AI, he said, can be remarkably insightful about what human beings need to do in order to become more humane.

There is much less evidence that anybody intends to take its advice.

The Questions Are the Story.

We began with a thousand words of rubbish.

We moved through humour, culture, logic and the importance of prompts. We asked whether machines should obey us or challenge us. We considered whether morality can be calculated, whether individuals are being asked to carry responsibilities that belong to institutions, and whether AI could help us become better people.

From there, we entered questions of work, value, power, parenthood and confidence in the future.

This was not a failure to remain on the subject.

It was the subject.

AI is not a narrow technological development that can be contained inside a discussion about software. It reaches into our ideas of intelligence, creativity, ownership, morality, work, responsibility and hope.

That is why every answer seems to produce another question.

At the end of our conversation, I returned to faith in humanity.

We may get things wrong. We may misuse the technology. We may follow it down impressive rabbit holes and mistake its confidence for wisdom.

But we must retain some faith in our own capabilities.

Humanity has muddled through before.

That does not guarantee that we will do so again. Bruce was quick to remind me that we have learned far more about technology than we have about ourselves.

Perhaps that is precisely where AI becomes most interesting.

It may not teach us how to be human by giving us a final answer. It may teach us by showing us our humour, contradictions, prejudices, ambitions, fears and possibilities from angles we have never been able to see before.

Technology is becoming a mirror.

Whether we learn anything from the reflection remains a human decision.