If you only look at Bill Gates’ talks about AI in recent years,
you might easily get one impression:
He is an optimist about AI.
He has compared the importance of AI
to that of:
personal computers,
the internet,
and smartphones.
He has also consistently believed
that AI has the potential to improve:
healthcare,
education,
agriculture,
global health,
and even reduce the gap between rich and poor countries.
So, in 2023,
when over a thousand tech figures called to pause the development of stronger AI,
Gates did not join them in saying:
"Stop everything."
His stance was practical at the time.
If you ask one group of companies to stop,
would others stop as well?
If the US stopped,
would other countries follow?
If the whole world cannot synchronize,
just asking a few companies to pause
might not solve the real problems.
But more than three years later,
Gates’ tone has clearly shifted.
Gates Is Now Seeking to Discuss AI with Xi Jinping
Reuters reported on August 26 that
Bill Gates is aiming to meet again with Chinese President Xi Jinping later this year.
A spokesperson said
the team is arranging a meeting during Gates’ next China trip,
tentatively scheduled for November,
though it is not yet formally confirmed.
What Gates wants to discuss
is not asking China to stop developing AI,
but rather:
If AI models become strong enough to pose global security risks, can the US and China establish joint rules?
His idea is that if the US imposes restrictions on models that might cause serious harm,
China might also be willing to take similar actions.
This represents a major shift.
Because the most common narrative about AI competition now is:
The US cannot lose to China.
China cannot lose to the US.
Everyone fears falling behind.
But Gates is now asking:
Are there some things that neither side can decide simply by "who does it first"?
He Has Not Suddenly Become Anti-AI
This is the easiest point to misunderstand.
Gates is not saying:
AI is too dangerous,
so we should stop developing it.
On the contrary,
his foundation still plans to invest substantial resources into AI applications in healthcare and global health.
The Gates Foundation expects to spend about $200 billion by 2045,
and AI is becoming an increasingly important part of that.
The foundation is researching:
how AI can assist prenatal health checks,
help healthcare personnel,
and bring expensive specialized expertise to resource-limited areas.
So Gates’ core stance is not:
Stop AI progress.
But rather:
As AI advances, society must also prepare itself.
What Really Changed His Tone Is the "Speed" of Progress
In Gates' latest long article, there’s a key sentence.
He says,
he has spent most of his life wishing innovation would move faster.
But with AI,
his feelings have become more complex.
The reason is simple.
Previously, when a new technology appeared,
society usually had some time to adjust.
When personal computers first came out,
not all jobs changed overnight.
Companies had to buy equipment.
Software had to be slowly developed.
Employees had to learn.
Workflows had to be redesigned.
The entire transformation could take ten to twenty years.
But AI is different.
Today, everyone already has:
phones,
computers,
and internet access.
Once new AI capabilities come out,
they can be globally adopted within days.
There’s no need to rebuild entire hardware systems,
nor to learn new programming languages.
You just talk to it in natural language.
So this time,
the pace at which technology changes jobs could be much faster than the speed at which governments revise systems.
His First Concern Is Job Displacement
In the past, when talking about AI replacing jobs,
people often said:
Agricultural mechanization caused unemployment before.
Factory automation caused unemployment too.
New jobs will eventually emerge.
That may be true.
But Gates is now worried:
Will this transition be so fast that people won’t have time to find new jobs?
In the past, an industry might shrink over 20-30 years.
The next generation could prepare by choosing other career paths early on.
But if AI alters
customer service,
accounting,
software development,
legal work,
administration,
and manufacturing
all within 5-10 years,
this won't be a gradual industry shift.
It might be a large-scale job redistribution.
This is why Gates has started to propose an idea unlike traditional tech optimism:
Should some jobs be deliberately reserved for humans?
“Human Reserved”: Not Because AI Can’t Do It, But Because We Choose Not To Give It Completely
Gates calls this concept:
Human Reserved.
You can think of it as a "human reserve zone."
Like a nature reserve.
Not because we can’t build on it,
but because even though we can,
we decide not to develop all of it.
Gates gives the example of care work.
His father suffered from Alzheimer’s in later years,
needing long-term care and companionship.
One day,
robots might be able to:
deliver meals,
remind about medication,
assist mobility,
monitor vital signs,
and even offer comforting words naturally.
Technically, this might be feasible.
But Gates believes some truly important aspects of caregiving
remain irreplaceable human values.
The same issue applies to:
education,
mental health,
medical communication,
and childcare.
The question is no longer just:
“Can AI do it better than humans?”
But rather:
Even if AI can do it, do we still want a human to be there?
In an interview with Reuters,
Gates even suggested discussing reserving up to 40% of existing jobs for humans.
This is not a finalized policy,
but a proposal to open social discussion.
Even More Surprising: He Also Proposes Taxing AI Tokens and Robots
Gates brought up this idea a long time ago,
but now has revived it.
His reasoning is straightforward.
Suppose a company hires an employee.
The company pays:
salary,
and payroll-related taxes,
and the government gains revenue from labor earnings.
But if the company buys a machine
to replace that person,
the machine is often considered a business investment or expense.
From a tax perspective,
sometimes:
Hiring a person costs more tax, while using a machine may save costs.
Gates thinks this may accelerate companies’ AI and robot substitution.
So he suggests,
in the future, tax could be considered on:
AI tokens,
robots,
or income generated by AI replacing labor.
What would the revenue be used for?
Retraining displaced workers.
Strengthening social safety nets.
Supporting workers with income losses.
Helping communities most affected by AI.
This is not about:
“Punishing AI.”
Rather, it hopes that:
after AI creates massive productivity gains, part of that benefit can fund the transition costs.
His Second Major Concern Is That Safety Risks Are Already Emerging Beyond Labs
If AI only:
writes a wrong email,
miscalculates numbers,
or generates a strange image,
we typically still have the chance to correct it.
But Gates’ current focus is at a different level.
For example:
Can AI assist in cyberattacks?
Can it make it easier for people without expertise to design dangerous biological threats?
Can it help design specific molecules?
Reuters reports that Gates wants countries to jointly monitor AI progress in
molecular design,
biological attacks,
and other high-risk capabilities.
His idea is not:
to ban AI from all biological research,
because AI might also help:
drug development,
disease research,
agriculture,
and vaccines.
What truly needs monitoring is:
when a beneficial capability crosses into dangerous territory.
This Is Why He Proposes a "Global AI Agency"
If only the US establishes AI safety rules,
but other countries do not,
what would happen?
Models can flow across borders.
Research results can spread internationally.
Cyberattacks can be launched from other countries.
Biological risks won’t respect national borders.
So Gates believes:
One country alone cannot govern AI effectively.
He now advocates for
a new international AI organization.
Not to directly copy any existing body,
but to borrow experiences from various international systems,
such as:
international nuclear inspections,
shared aviation safety rules,
and agreements to protect the ozone layer.
These systems share a common trait:
they address problems crossing borders,
and no single country alone can solve them completely.
AI is beginning to exhibit the same characteristic.
Why Are the US and China Unavoidable Participants?
Because the most advanced AI
is highly concentrated in a few countries and companies.
If a small country signs an impressive AI pact,
but countries controlling:
the strongest models,
largest computing power,
advanced chips,
and major AI companies
do not participate,
the pact’s impact is extremely limited.
So Gates especially points out:
The US and China must cooperate to some extent.
This does not mean the two countries must stop competing,
nor that they will suddenly trust each other.
But some risks may present a realistic common interest:
I don’t want your AI to cause a global disaster, and you don’t want mine to cause one either.
Similar to nuclear safety.
Two countries can be rivals,
yet still agree:
certain accidents must never happen.
Does This Contradict Gates’ 2023 Position?
At first glance, it might seem so.
In 2023:
People called for an AI pause.
Gates said:
it’s impossible for the world to pause together.
In 2026:
Gates himself calls for global governance.
But if you look closely,
both positions can coexist.
What he opposed then was:
"Asking everyone to stop developing technology."
What he calls for now is:
"Technology keeps advancing, but dangerous capabilities, social impacts, and cross-border risks must be managed."
The simplest analogy is:
Previously, someone might say:
"Cars are getting faster and faster, let’s stop making cars for now."
Gates found that unrealistic.
Now his thinking is more like:
"Okay, cars keep being made. But since they’re already this fast, shouldn’t we create driver’s licenses, traffic lights, guardrails, and safety inspections together?"
Not stopping progress,
but:
once progress reaches a certain point, the rules must also be upgraded.
What Really Changed Is Not Gates’ Expectations for AI
He still believes AI can do a lot of good.
He might even trust its capabilities more than before.
But because AI has become so powerful,
he now worries more about:
what if we use it wrongly?
What if benefits concentrate in the hands of a few?
What if job loss outpaces retraining?
What if dangerous capabilities cross borders?
The real turning point is not:
"Bill Gates going from believing in AI to fearing AI."
But rather:
"Bill Gates going from believing AI will change the world, to asking: is the world ready to handle that change?"
This might be the biggest AI debate in the coming years.
We have spent much time asking:
Which model is strongest?
Which agent is smartest?
Which robot is most humanlike?
The next, tougher question might be:
When AI advances faster than systems can keep up, what should we reserve for humans?
Jobs?
Responsibility?
Safety?
Privacy?
Or final decision-making power?
Tech companies will keep making AI stronger.
What is not done yet
is for humanity to decide:
At what point of advancement do new rules become necessary.
Today, advance alongside AI.
Learn one AI skill a day.
Save a bit of time every day.
Improve your ability little by little each day.
SasaDaily, growing with you.
Recommended Reading
Why Did South Australia Partner with OpenAI and Then Turn Around to Create an AI Royal Commission?
Why Do Bosses Expect You to Do More Even Though AI Saves You Work Time?
