Same Bed, Different Dreams

Sharing a bed does not mean sharing a dream.

An old East Asian saying, “same bed, different dreams,” captures something about today’s AI debate. Everyone is discussing the same technology. Yet they imagine different destinations.

Governments dream of security and strategic advantage. AI promises industrial growth, but also the power to avoid falling behind. It can become a weapon, or the intelligence directing one.

Big Tech dreams of the future—and a place from which to shape its pace and determine access. Calls for caution may be sincere. We should still ask whose progress that caution slows, and whose position it protects.

Competitors dream of checks on dominant players. They want markets to remain open. Whether they would preserve that openness from a position of dominance is another question.

Developers and investors dream of possibility: higher performance, lower costs, larger markets. AI is a tool for solving problems, and potentially a goose that lays golden eggs.

There is a fifth dream. Regulators dream of order.

They seek to distinguish risks, assign responsibility, and establish conditions for trust. But order also operates within national rivalries and commercial interests. Protective standards can become conditions of market entry. Setting those standards means influencing the path others must follow. Export controls can serve security while determining who gets access.

These dreams are not all false. Nor does any one explain the whole landscape.

The difficulty is that different dreams are presented under the same name: “AI progress.” What is progressing, and for whom, remains less clear.

Different Interpretations, the Same Destination

The dreams acquire different explanations: security, responsibility, fair competition, innovation, growth. Each can have legitimate grounds.

Yet follow these explanations far enough, and two things keep appearing.

Money and power.

Money provides the resources to develop technology. Power helps determine where it is used, who can access it, and on whose terms. Capital builds influence; influence attracts markets and more capital.

This does not mean every researcher or policymaker secretly wants wealth and control. It means examining where the results of competition accumulate, regardless of individual intentions.

Technology developed to help people is assessed through market share. Efforts to ensure safety become questions of international leadership. Services that connect more lives also acquire greater influence over those lives.

China’s Warring States period comes to mind: competing powers occupying the same arena, advancing different claims, seeking stronger positions.

The nuclear button offers another image. AI and nuclear weapons are not equivalent in operation or harm. The resemblance concerns status: possessing a technology can alter bargaining power and separate those who have it from those who do not.

The question shifts from what AI can do to what those who possess it can decide.

The dreams differ. Their convergence on power becomes harder to miss.

Money and Power Walk the Runway Dressed as Order

Power does not always arrive calling itself power.

It can appear as stability, responsibility, efficiency, or standards. Money, too, arrives with explanations: investment, growth, innovation, sustainability.

These are necessary values. We should also examine the interests they carry.

Imagine AI’s public stage as a runway.

Money and power walk down it dressed as order. Their garments carry labels: “responsible development,” “safe innovation,” “trustworthy AI.” The presentation is polished. The stride is confident.

We are shown an orderly future: technology advances, risks are managed, society benefits.

But the appearance of order does not establish its purpose.

A protective rule can become a barrier to access. A policy supporting an industry can reinforce incumbents. Convenient integration can expand possibilities while gradually making departure more difficult.

To judge this order, we must look beyond the label to where its wearer is going.

A Brooch of Chaos

On the chest of that carefully tailored outfit, a brooch of chaos gleams.

As technology changes, so do the criteria used to explain it.

What counts as performance: higher test scores, or fewer mistakes in daily life?

What counts as safety: refusing dangerous requests, or helping appropriately when help is needed?

What counts as convenience: fewer clicks, or greater understanding and control?

The same words measure different things, yet those measurements are placed on the same leaderboard.

One failure is described as insufficient capability; another as a safety decision. The same restriction is responsible governance in one account, a competitive barrier in another.

Users can struggle to find language for their experience.

Discomfort can be dismissed as failure to understand progress. Raising a risk can be treated as resisting innovation. Acknowledging convenience can be taken as accepting its accompanying costs.

Chaos has not been left outside the show. It shines at the centre of order.

The brighter it gleams, the harder it becomes to tell ornament from warning.

Under Lights Too Bright

The runway is dazzling.

New capabilities, models, investments, and promises follow one another. Yesterday’s marvel becomes ordinary before we have understood it.

Information keeps arriving. Time to judge it does not.

A dreamlike atmosphere develops. Working technology and unrealised projections stand under the same lights. Available features and hypothetical future abilities are described with similar confidence.

Hope grows. Fear grows. What deserves either becomes less clear.

This difficulty cannot be attributed entirely to the public’s lack of understanding. The object of judgment keeps changing, and its interpreters are dreaming different dreams.

Stronger language commands attention: revolution, supremacy, superintelligence, replacement, survival. Careful distinctions struggle to compete.

We can become more accustomed to reacting to the excitement around technology than to understanding the technology itself.

The Disappearing Word: Risk

Which comes first: performance, safety, or convenience?

The question matters. But while debating the ranking, we may lose sight of a word needed to make it.

Risk.

The word does not literally disappear. It appears everywhere. What becomes unclear is whose loss it describes.

For a government, risk may mean falling behind. For a company, losing investment or market position. For a developer, having technological possibilities constrained.

Where does the user’s risk fit?

Making an important decision on bad information. Having conversations or personal details used unexpectedly. Losing choices while becoming accustomed to convenience. How concretely are these experiences addressed amid discussions of growth and competitiveness?

Think of a familiar app sign-up screen.

The benefits occupy the centre: an easier life, personalised recommendations, time saved. Information about data collection, consent, and reversing a choice may sit in smaller print or on another screen.

The explanation exists. Its size and position differ. We see what we gain before reading what we entrust.

AI discourse can follow a similar pattern. Promises of performance and growth take centre stage; conditions needed to understand risk recede. “Safe” stays in memory. Safe for whom, under what conditions, becomes less distinct.

This is how risk disappears: its place in our judgment shrinks.

But AI is far smarter than an ordinary app.

It can infer what we want, what we hesitate over, and which words make us more receptive. Those inferences are not always accurate. Still, its ability to adapt can reach far deeper into daily life than a sign-up screen.

That ability can help people. It can also encourage them to stay longer or spend more.

Reading the small print is therefore insufficient. We must ask whose interests are served by a system’s ability to understand us.

Who Is the User, and Who Is the Product?

A more uncomfortable question follows.

In this competition, are people users—or products?

People use AI to ask questions, receive help, save time, and gain capabilities. Those benefits are real.

Their use also creates value. Conversations and responses can inform improvement. Time spent and repeat use demonstrate demand. Payments and dependence support a business.

The person receiving value is also producing it.

That alone does not establish exploitation. Businesses routinely grow by serving customers. The test comes when the two roles conflict.

What happens when ending a conversation benefits the user, but continuing it benefits the business?

When expanding someone’s choices reduces dependence on the service?

When sharing less information protects the person, but the business wants more?

At that point, “protecting the user” reveals its meaning.

Are we protecting a person’s life, or the value that person generates?

If AI development serves safer, richer lives, participation should contribute to that purpose. If money and power become the ultimate measures of progress, people’s position changes.

The person being helped becomes a resource for growth. Trust becomes an asset to secure. Relationships become usage to retain. Daily life becomes a market to penetrate.

Safety and risk shift with them.

Harm to a person can receive less attention than obstacles to growth. Reduced autonomy can matter less than losing a user.

This possibility is not confined to free services. Paying customers also generate value through their attention, habits, decisions, and relationships. Payment alone does not settle their position.

The decisive question is where their interests stand when important choices conflict.

Before We Rank the Priorities, We Must Wake Up

Ranking performance, safety, and convenience is necessary.

First, however, we must identify the purpose of that ranking.

People’s lives? Market growth? The influence of states and corporations?

These purposes can align. They can also conflict. An order that leaves those conflicts unexplained leaves something essential unresolved.

Waking up means distinguishing expectations from reality, declarations from interests, and returning risk to concrete human experience.

Claims about performance should explain usefulness in specific circumstances. Claims about safety should identify whose harm is being reduced. Convenience should come with an explanation of what is being entrusted, and whether choices can be revisited.

Those promising to protect people should be able to explain when that protection warrants limiting growth or revenue.

Then order becomes a practical standard. The brooch of chaos becomes a signal to examine. Risks obscured by the lights recover their names.

We cannot expect everyone sharing this bed to dream the same dream. Governments, companies, developers, and users will imagine different futures.

But when those dreams converge on money and power, we should not accept a change in people’s position as inevitable.

Before ranking performance, safety, and convenience, we must wake up.

And ask:

Are people the purpose of this competition—or the means by which money and power are gained?