When Information Becomes Cheap, Curiosity Becomes Valuable.

Bruce Lloyd, London and Frank Peters, Cleebourg.

AI can produce more information than any of us can absorb. The harder question is what we choose to do with it — and whether asking better questions may matter more than having more answers.

I had been watching Bruce’s output for some time, and there was something about it I could not quite understand.

One morning there would be AI and the future of learning. Then the United Nations. Then H. G. Wells and Keynes. W. H. Auden. The Old Masters. Star Wars. Marcel Proust and James Joyce. Then, with no obvious warning, a North London branch of Waitrose and what it might tell us about multicultural society.

There seemed to be no pattern.

So I asked him.

“There’s a high element of randomness,” Bruce said.

That was reassuring. At least I had not missed some sophisticated organising principle.

But the randomness turned out to be more interesting than it sounded. There were books on his shelves he felt he had never done justice to, things in the news that caught his attention, and what he described simply as “an insatiable curiosity across all sorts of areas.”

Something appears. It raises a question. The question connects with something else. Bruce gives it to AI and sees where the conversation goes.

Nobody has commissioned the work. There is no market waiting for it and, by his own admission, much of it attracts very little attention. He calls it his “creative hobby”.

The more we talked, the more that small phrase began to seem rather important.

Following the question.

Bruce had recently been listening to Julian Jackson’s substantial biography of Charles de Gaulle, A Certain Idea of France.

Listening matters here.

He had tried reading the book and found himself doing what many experienced readers probably do with a large volume: looking for the argument, searching for the essence, trying to establish quickly what the author was really telling him. Years of reviewing books had trained him to extract.

The audiobook changed that. He listened while moving around the house, shopping or clearing unimportant emails, and found himself staying with details that he might otherwise have skimmed past.

That immediately produced another question: what do different ways of receiving information do to the way we learn? And then another: what happens when AI enters that process?

This is how Bruce’s subjects tend to develop. One question does not produce an answer and finish the job. It opens a door to the next one.

He had also been thinking about C. P. Snow’s idea of the “two cultures”, the divide between scientific and humanistic intellectual life, and wondering what happens to such divisions in an age when specialist information can be summoned almost instantly.

That led him towards what he called a potentially strange new specialism: “the capacity to integrate”.

Perhaps the increasingly useful person is not simply the one who knows more and more about less and less. There may also be growing value in somebody who understands enough of several specialist territories to see how they connect.

AI changes the equation because access to specialist information is becoming extraordinarily easy.

Knowing what to do with it is rather less so.

An impossible question in Waitrose.

The Waitrose experiment began almost accidentally.

Bruce was in a busy North London supermarket and found himself wondering how many nationalities might pass through the shop during an ordinary day, among both customers and staff.

It was, as he put it, “an impossible question”.

So naturally he asked ChatGPT.

There was no secret database containing the nationalities of Tuesday afternoon’s shoppers, and AI could not suddenly discover one. What interested Bruce was what happened next. Faced with a question it could not answer directly, the system began offering information around it and trying to construct something useful from what was available.

That led to a broader experiment. Why might a highly multicultural collection of people function together perfectly normally inside a supermarket? Could that tell us anything about multicultural society more generally?

I immediately complicated matters by pointing out that Waitrose itself might distort the picture. Where are its stores? Who shops there? Would Tesco produce a different result?

Bruce had already begun pushing in much the same direction.

Then something smaller happened which fascinated him even more.

He did not like the word “teach” in a proposed question about what Waitrose could teach us about multicultural society. He preferred “learn”.

It looked like a tiny edit.

But according to Bruce, ChatGPT did not simply swap one word for another. The change substantially altered the direction of the analysis.

For him, that was the interesting part. The value was not simply in whatever finished text appeared at the end, but in seeing how the thinking changed when the question changed.

“A lot of these exercises,” he told me, “are not because of just the end content, but what they tell us about the process that AI goes through in order to get to that content.”

That is a rather different use of AI from treating it as a machine into which we insert a question and from which the Answer emerges.

The result matters.

But so does watching the question move.

Credit, benefit and a small audience.

At one point I asked Bruce who deserved the credit for all this.

Was it Bruce Lloyd, choosing the subjects, making the connections and continuing to ask questions? Or the collection of AI systems helping him explore them?

Bruce changed one word in my question.

Perhaps, he suggested, the better question was not who gets the credit?

It was who gets the benefit?

Bruce certainly does.

“It keeps my mind going. It keeps me relatively active. And it is what I would like to call a creative hobby.”

Other people may benefit too, if they happen across something he has produced and find it useful. Occasionally somebody comments. Occasionally somebody responds.

But this is not a mass-audience enterprise and Bruce is under no illusion that it is.

I pushed him on that. He can spend time exploring a question, produce another piece of work, place it in the public domain and then watch the counter move upwards very slowly. Perhaps a few people read it. Perhaps one leaves a comment.

Meanwhile, he has satisfied his own curiosity and learnt something.

Why isn’t that enough?

“In a way it’s enough,” he said. “In a way, it’s more than enough if just one person comes back and says that they appreciated it.”

There is something wonderfully unfashionable about that answer.

We have become accustomed to measuring intellectual output by readership, views, followers, engagement and reach. The machinery quietly encourages the belief that something unread has somehow failed.

Bruce’s work already has value before anybody else arrives.

The audience is a bonus.

When information becomes cheap.

Our conversation kept circling back to one uncomfortable fact.

Information is becoming incredibly cheap.

AI can produce enormous amounts of it, very quickly, on almost any subject. Text, analysis, summaries, comparisons, explanations, research starting points, images and endless variations on all of them.

I wondered whether that made the human contribution less valuable.

Bruce thought the opposite.

“The more information you have,” he said, “the more important humans become in order to decide what it is that should be done with that information.”

That may be one of the more interesting tensions underneath the AI debate.

We spend a great deal of time concentrating on what the systems can produce. But abundance creates its own difficulty. If I can generate fifty analyses before breakfast, why should I read any of them? If everybody can produce information, what makes a particular piece worth paying attention to?

Relevance begins to matter more than volume. Judgment becomes more important than retrieval. Curiosity matters because somebody still has to decide where to look and why.

And Bruce’s slightly odd new specialism — integration — begins to make more sense.

None of this means AI gives us absolute answers. Bruce repeatedly returned to that qualification. At best, he sees it as helping to create “a better framework within which to take decisions”.

The decision still belongs elsewhere.

Questions it cannot answer.

One of Bruce’s habits stayed with me more than most.

He likes asking AI questions he already suspects it cannot answer.

At first that sounds slightly perverse. Surely the point of an information system is to ask it questions for which an answer exists?

Bruce is interested in what happens when one does not.

If the system cannot answer the precise question, what does it substitute? What does it decide might be relevant? Which neighbouring information does it reach for?

The failure becomes part of the experiment.

It also teaches the human questioner something about the tool.

This is very different from passively accepting polished output because it looks convincing. Bruce is poking at the edges, changing words, asking awkward questions and watching what shifts.

The machine responds.

The human wonders.

We wandered through learning, education, academic research, responsibility, technology and the ability of institutions to make simple questions remarkably complicated. At several points we nearly disappeared down entirely different roads.

Yet the distinction kept returning.

AI can help us process more, explore faster and discover connections we might not have noticed alone. It can provide frameworks and bring unfamiliar territory within reach.

What it cannot decide is why a particular question matters to us in the first place.

The Waitrose question.

Towards the end of our conversation, I asked Bruce to choose his “pick of the week” from his recent output.

He hesitated. They were all interesting for different reasons.

Eventually he chose Waitrose.

Not because it was the grandest subject. Quite the opposite.

It began with an ordinary observation in a supermarket and a question that could not really be answered. Then one word changed — teach became learn — and the direction of the enquiry changed with it.

Perhaps that is why I like the story too.

For all the enormous claims made about AI, this version feels much more human.

You notice something ordinary and become curious about it. You ask a question. The answer is imperfect, so you alter the question. Something unexpected appears and you look again.

Perhaps another person eventually sees what you have done and finds it interesting too.

Perhaps nobody does.

Bruce would probably continue anyway.

In an age capable of producing more information than any of us could possibly consume, that may be part of where the value now lies.

Not in having access to another million answers.

In having a reason to ask the next question.

Bruce’s Pick of the Week.