权力与算力同桌：G7 的 AI 午餐会，和一场没有结论的较量

Today we'll talk about something that seems ordinary but is actually deeply significant.

On June 17, in the small French town of Évian,

on the final day of the G7 summit,

a special working lunch was held.

Let's first lay out the background.

G7 — the 'Group of Seven' —

is made up of the US, UK, France, Germany,

Japan, Italy, and Canada,

these seven developed countries.

Every year, these countries' leaders

gather together

to discuss the world's most important issues.

And this year, the most-watched issue,

with no suspense, is artificial intelligence — AI.

What truly made this lunch special was its seating.

On one side of the table

were the heads of state who hold political power,

including US President Trump

and French President Macron, among others.

On the other side

sat a group of people who hold technological power —

the leaders of today's top global AI companies.

According to reports, nearly a dozen tech leaders attended,

including OpenAI's Altman,

Google DeepMind's Hassabis,

Anthropic's Amodei,

plus executives from Meta, Mistral, Cohere,

Salesforce, and others.

You could say half of global AI

was seated at this one table.

The theme of this lunch,

in the official wording, was:

'Ensuring artificial intelligence is deployed safely, rapidly,

and effectively.'

Put plainly, the people at the table

were all trying to answer the same question:

when AI is this powerful,

how do we steer it?

The topics discussed were wide.

From the most frontier AI risks,

to the infrastructure that supports AI,

to a country's 'tech sovereignty',

and even how to protect children online.

But the core tension of the whole lunch

was concentrated on one question:

who, in the end, sets the rules for AI?

On this point,

several American tech-company leaders

gave a fairly consistent direction.

Anthropic's Amodei,

and Google DeepMind's Hassabis,

both called for building

a US-led AI coalition —

with America taking the lead, uniting a group of countries,

to jointly set AI's standards and rules.

OpenAI's Altman

went a step further, proposing

an international forum,

to give each country's AI

a unified, impartial test of capability and risk,

and to be a platform for cooperation among nations.

These proposals sound constructive.

But if you stand in Europe's shoes,

the flavor is a little different.

In recent years, European countries led by France

have grown more and more wary

of America's absolute dominance in AI.

They repeatedly stress one concept:

'tech sovereignty'.

What they imply is:

Europe is unwilling

to passively accept America's technology,

America's standards, the game rules America sets.

France stated clearly

that what Europe wants is the ability to compete,

not merely to comply and follow.

So you see,

this seemingly warm and friendly lunch,

beneath the table,

was a quiet contest over

future say-so.

On one side, the creators of the technology,

hoping that they, and their own country,

lead the making of the rules.

On the other, other countries,

unwilling to hand their future

entirely into someone else's hands.

So, this lunch —

did it ultimately produce a result?

The answer is: almost none.

Afterward, no specific policy was released,

and no binding agreement was formed.

Everyone just sat down

and put their positions and worries

on the table — that was all.

But, I'd say,

don't underestimate such a 'resultless' lunch.

Because it itself

sends a clear signal:

artificial intelligence

is no longer just a matter for tech companies.

Its importance

has risen to the point

where it needs the world's most powerful political leaders,

and its most capable technological leaders,

to sit down face to face

and discuss it in earnest.

This meal

is a microcosm of this era.

At the end of the show,

I want to talk with you about

the more fundamental question behind this.

In human history,

every sufficiently powerful new technology —

from gunpowder to nuclear energy —

has faced the same question:

in whose hands should its steering wheel be held?

In the hands of those who understand it best,

and are most driven to push it forward?

That might be most efficient,

but would it lack restraint?

Or in the hands of more people,

more countries?

That might be safer and fairer,

but would it be too slow, too chaotic?

AI

is exactly the new 'fire'

our generation must face.

This time,

can we

do a little better than before?

So finally,

I'll leave you with two questions.

First, whom do you trust to set the rules for AI?

Tech companies, governments,

or some entirely new way where everyone takes part?

Second, back to yourself.

In this era of AI changing so fast,

what is it that you hope

to hold firmly in your own hands?

These two questions have no standard answer.

But they are worth thinking about slowly.