The Hallucinogenic Confidence Trick: AI, All That Jazz, and Why It Might Be Time to Say Goodbye

Every great confidence trick begins with something that looks real.

The victim doesn't hand over their wallet because they are stupid. They hand it over because the illusion is convincing. The suit looks expensive. The story sounds plausible. The paperwork appears authentic. The confidence is irresistible.

The trick works because confidence is often mistaken for truth.

That observation should make us deeply uncomfortable when discussing artificial intelligence.

For the past few years we have been told that AI represents the future of knowledge work. It will write our reports, build our applications, create our strategies, answer our questions, and eventually make better decisions than the people who created it. The demonstrations are extraordinary. The outputs are often impressive. The productivity gains can be genuine.

Yet something about the entire conversation feels increasingly strange.

The more capable these systems become, the less certain I am that we are talking about intelligence at all.

What if we are simply witnessing the most sophisticated confidence trick ever created?

Not a deliberate deception. Not a conspiracy. Something far more dangerous.

A machine that can sound right regardless of whether it is right.

For centuries, knowledge and confidence travelled together because they were expensive. Expertise required years of experience. Authority had to be earned. Evidence took time to collect. The process was slow, frustrating and imperfect, but it contained an important safeguard. Confidence was constrained by effort.

Artificial intelligence has broken that relationship.

For the first time in history, confidence can be manufactured instantly.

Ask a question.

Receive an answer.

Ask for a strategy.

Receive a strategy.

Ask for an architecture.

Receive an architecture.

Ask for a business plan.

Receive a business plan.

The machine never hesitates. It never doubts itself. It never pauses to wonder whether it has misunderstood the objective. It simply continues.

And therein lies the trick.

We have spent years describing AI as though it were a thinking machine when it behaves far more like a jazz musician.

A jazz musician hears a tune and carries it forward. One phrase leads to another. One riff becomes the foundation for the next. The performance evolves through continuation.

Sometimes the music is beautiful.

Sometimes the music is terrible.

The musician's talent lies in recognising patterns and extending them.

The musician does not define the destination.

Current AI behaves in much the same way.

It hears the tune and keeps playing.

The problem is that organisations have started mistaking continuation for understanding.

A report appears.

A recommendation appears.

A roadmap appears.

A thousand lines of code appear.

The machine sounds confident because confidence is built into the act of continuation. The next note must follow the previous note. The next sentence must follow the previous sentence. The next idea must emerge from the previous idea.

The performance never stops.

Nobody notices that the musician has absolutely no idea whether the song is worth playing.

This is where the internet itself is beginning to change.

The early web was messy, chaotic and frequently wrong. Yet it possessed a quality that now feels increasingly rare. Most content had a human author standing behind it. There was accountability. There was intent. There was at least some connection between the words and the lived experience that produced them.

Today the internet feels different.

It feels padded.

Bloated.

Artificially inflated.

Pages appear that nobody really wrote. Articles appear that nobody truly believes. Software appears that nobody completely understands. Entire businesses emerge around content that exists solely because a machine was capable of generating it.

The web is becoming a hall of mirrors built from continuations.

Every article references another article. Every summary references another summary. Every generated explanation references another generated explanation. The machine consumes the output of yesterday's machine and produces the output of tomorrow's machine.

The tune continues.

Reality quietly leaves the room.

The technology industry insists that this represents progress.

Perhaps some of it does.

Yet there is a growing suspicion that we are witnessing a form of digital dry rot.

Not visible decay.

Structural decay.

The websites still load.

The applications still run.

The reports still look professional.

The presentations still receive applause.

Everything appears intact.

The problem is that nobody can quite remember where the foundations are.

That is the danger of hallucinogenic confidence.

Not that the machine lies.

Not that the machine is malicious.

But that the machine can produce an endless stream of plausible continuations which slowly replace the need for verification.

Why check the facts when the report looks convincing?

Why inspect the code when the application appears to work?

Why challenge the strategy when the slide deck is beautifully written?

The confidence becomes the evidence.

And that is where the confidence trick completes itself.

The final stage of every confidence trick is not deception.

It is participation.

The victim begins helping the illusion sustain itself.

Businesses are increasingly doing the same thing.

We measure AI adoption.

We measure AI usage.

We measure AI-generated output.

We celebrate the volume of activity.

The machine keeps playing.

The audience keeps clapping.

Nobody asks whether the song should have ended twenty minutes ago.

Perhaps that is the question we should be asking now.

Not whether AI will become more intelligent.

Not whether models will become larger.

Not whether agents will become more autonomous.

A simpler question.

Has the performance run its course?

Because there comes a point in every jazz session when the greatest contribution a musician can make is not another note.

It is silence.

The technology industry appears convinced that the answer to every problem is more AI, more content, more code, more agents, more automation, more continuation.

History suggests otherwise.

Most mature technologies eventually discover that value does not come from doing more.

It comes from knowing when enough has been reached.

Artificial intelligence may prove enormously useful.

It may transform industries.

It may change the nature of work.

But if we continue to mistake confidence for truth, continuation for understanding, and generation for value, then we risk building an internet that sounds increasingly impressive while becoming progressively detached from reality.

And that is the oldest confidence trick of all.

The audience leaves believing they witnessed something extraordinary.

Long after the magician has packed up and gone home.