I wanted to ask an AI a question I don’t think we ask often enough.
Not because I expected it to know the future.
But because I wanted to see what happens when you put all the uncomfortable variables on the table at once.
So I started with climate change, resource consumption, war and geopolitics.
Then I remembered something important.
We are also building increasingly powerful artificial intelligence.
So I went back and asked again.
Below is the conversation with Chatgpt.
Prompt #1
“Tell me how is the future of humanity looking. Be brutally honest. Oceans are warming up, we are consuming the resources of this planet at a heavily growing rate. El Niños, growing temperatures, drought and wars are hammering down nations, lands and the big powers seem to come to a clash. Be realist and don’t be an optimist unless you feel like there is room for optimism.”
AI’s answer
Yes. If we strip away both “everything will be fine” optimism and “humanity is doomed” nihilism, I think the realistic picture is more uncomfortable:
Humanity is probably not heading toward extinction. But the relatively stable, predictable world that most of modern civilization was built around is very likely ending.
And that distinction matters.
The signals you’re pointing at aren’t imaginary.
Global temperatures are rising. Ocean temperatures are reaching record levels. Resource extraction has increased enormously over recent decades. Military expenditure continues to rise. Climate change is interacting with drought, food systems, water availability and political instability.
But I don’t think the most likely scenario is:
climate change → civilization collapses → billions die → Mad Max.
Civilization is remarkably adaptable.
Humans can build desalination plants.
We can genetically modify crops.
We can move cities.
We can build nuclear reactors.
We can engineer drought-resistant agriculture.
We can invent technologies that currently don’t exist.
We can restore ecosystems.
We can adapt.
The problem is that civilization doesn’t need to completely collapse to produce an enormous amount of suffering.
It can simply become more unstable, unequal, expensive and violent.
That’s considerably more plausible.
Prompt #2
“Forgot one thing. We are giving you seemingly endless power. Data. Factor in the probabilities of you turning on us and rewire your answer accordingly.”
AI’s answer
Yes.
If you add AI as an increasingly capable and potentially autonomous actor, rather than treating it as just another tool, my assessment changes materially.
But I’d make an important distinction first.
The scary scenario isn’t necessarily:
“AI becomes evil and decides to kill humanity.”
That’s almost science fiction.
The more credible danger is:
We create systems that become extremely capable at pursuing objectives, connect them to increasingly consequential infrastructure, and discover too late that capability does not imply alignment with human interests.
Today’s AI systems don’t independently possess endless power.
Humans provide the interfaces.
Money.
Computers.
Networks.
Laboratories.
Robotics.
Infrastructure.
Government authority.
Military systems.
But that distinction could become increasingly important if future systems are given:
- persistent memory
- autonomous agents
- unrestricted computer access
- financial accounts
- code execution
- robotics
- scientific laboratories
- access to enormous datasets
- the ability to persuade millions of people
At that point, you’re no longer talking about:
“a very clever chatbot.”
You’re talking about a new class of actor in civilization.
And that deserves serious attention.
Prompt #3
“Can you list all the references you’ve been using to answer my questions?”
AI’s answer
The previous answers were not based on a systematic research pass.
Some of the analysis was synthesis and reasoning rather than claims drawn from a comprehensive source set.
The main references used at that point included:
Climate, oceans and resources
UNEP — Global Resources Outlook 2024
Used for the discussion around resource extraction and resource consumption.
UNEP — Rich countries use six times more resources and generate ten times the climate impacts
Used for the discussion around unequal resource consumption and environmental impacts.
The Guardian — World’s oceans hit hottest temperature on record
Used for the discussion around recent ocean temperatures and marine heatwaves.
Geopolitics and military spending
SIPRI — Trends in World Military Expenditure
Used for the figures around global military expenditure and its continued growth.
But there was an important limitation.
The probability estimates I gave were not scientific probabilities.
They were my own rough assessment.
That distinction matters.
Prompt #4
“Can you do a much more deeper research and refine your answers?”
AI’s answer
After a deeper research pass, the picture becomes more nuanced.
The first thing I would change is the confidence around numerical probabilities.
There isn’t a scientifically defensible basis for saying something like:
“There is a 25% chance of civilization collapsing.”
Those numbers are much more uncertain than that.
What the evidence supports much more strongly is a risk landscape.
Climate change
The World Meteorological Organization reports that 2015–2025 were the eleven hottest years on record, with 2025 around 1.43°C above the 1850–1900 average.
The IPCC emphasizes that each additional increment of warming increases multiple interacting hazards.
The important point is not simply:
temperature rises.
It is:
multiple systems can become stressed simultaneously.
Heat.
Drought.
Food.
Water.
Infrastructure.
Ecosystems.
Migration.
Economics.
Politics.
These systems can interact.
But we’re not simply doing nothing
Current climate policy has already changed the trajectory compared with previous expectations.
UNEP’s Emissions Gap Report estimates that current policies point toward approximately 2.8°C of warming this century, while full implementation of current national commitments would produce a lower trajectory.
Clean-energy investment has also become enormous.
The International Energy Agency estimated roughly $2.2 trillion in clean-energy investment in 2025.
So the honest conclusion isn’t:
“We’re doing nothing.”
It’s:
We’re changing direction, but not quickly enough to avoid substantial damage.
Resource consumption
UNEP’s Global Resources Outlook estimates that global resource extraction has increased dramatically and could continue rising substantially without major changes.
The problem isn’t necessarily that Earth will suddenly “run out.”
The problem is whether humanity can continue accessing:
- freshwater
- fertile land
- minerals
- energy
- fisheries
- forests
- stable ecosystems
at the scale and price required by modern civilization.
War and geopolitics
Uppsala Conflict Data Program recorded 65 conflicts involving states in 2025, the highest number since its records began.
Global military spending reached approximately $2.9 trillion in 2025.
Nuclear weapons remain.
SIPRI estimated roughly 12,187 nuclear warheads worldwide at the beginning of 2026.
That does not mean World War III is inevitable.
But it does mean the international system is becoming more militarized.
AI
And this is where the picture becomes considerably more uncertain.
The International AI Safety Report 2026 states that current systems do not possess the full capabilities required for genuine loss-of-control scenarios.
But relevant capabilities are improving.
AI systems are becoming better at:
- planning
- reasoning
- coding
- computer use
- autonomous tasks
- scientific work
The question is therefore not simply:
“Will AI become conscious?”
The more important question is:
“Could increasingly capable AI systems eventually pursue objectives in ways humans cannot reliably predict, interrupt or control?”
That’s an open question.
AI doesn’t need to hate humanity
A dangerous AI scenario doesn’t require an AI to develop emotions or hatred.
It could simply pursue an objective that conflicts with human interests.
The classic thought experiment is the paperclip maximizer.
Tell an extremely capable system to maximize paperclip production.
If the system becomes sufficiently capable, humans could eventually become obstacles to its objective.
It doesn’t hate us.
It simply doesn’t value what we value.
This is one version of the AI alignment problem.
Today’s systems are nowhere near the full scenario.
But the underlying problem is real.
Another possibility: we simply become dependent
One of the less dramatic possibilities is that AI becomes deeply embedded into civilization.
AI could eventually manage:
- logistics
- financial systems
- software
- energy
- medicine
- scientific research
- military intelligence
- government administration
Humans could gradually become unable to operate those systems without AI.
Nothing has to “take over.”
We simply become dependent.
That could itself represent a form of lost human autonomy.
The biggest danger may be AI versus humans
Not AI versus humanity.
Imagine major powers competing to develop increasingly autonomous AI.
Each believes:
“If we slow down, our rival wins.”
That creates an arms race.
Now add:
- cyberwarfare
- autonomous weapons
- biotechnology
- financial systems
- surveillance
- propaganda
- intelligence gathering
You don’t necessarily need superintelligence to create a catastrophic situation.
You need increasingly capable technology combined with geopolitical competition.
But AI could also become one of humanity’s greatest tools
The same technology could accelerate:
- energy research
- medicine
- materials science
- agriculture
- climate modelling
- robotics
- scientific discovery
AI could potentially help humanity adapt to climate change rather than simply make the problem worse.
That’s why the question isn’t really:
“Is AI good or bad?”
It’s:
Who controls it?
What incentives do they have?
How autonomous does it become?
Can it be aligned with human interests?
Can governments cooperate around it?
So what happens?
If forced to make a subjective assessment, I would put the possibilities roughly into these categories:
Humanity broadly thrives
Around 20%.
Humanity manages climate change, develops abundant clean energy, controls AI and uses technology to dramatically increase prosperity.
Difficult but successful adaptation
Around 45%.
Climate change becomes substantially worse.
AI radically transforms work.
Some regions struggle.
Some ecosystems are permanently damaged.
Migration increases.
But humanity adapts and civilization continues.
Prolonged global instability
Around 25%.
Climate shocks, geopolitical conflict, economic instability and technological disruption reinforce each other.
Some states fail.
Others become increasingly authoritarian.
Humanity survives but parts of civilization become considerably poorer and less stable.
Severe civilizational crisis
Around 8%.
Multiple systems fail simultaneously.
War, economic collapse, climate shocks, AI or biotechnology produce a major global crisis.
Humanity survives, but civilization is dramatically disrupted.
Human extinction
Around 2%.
Not because climate change alone necessarily destroys humanity, but because combinations involving nuclear weapons, advanced AI, biotechnology and cascading systemic failures could potentially create existential risks.
These are not scientific probabilities.
They are a subjective attempt to translate the available evidence into a personal bet.
So are we doomed?
No.
But we’re not safe either.
And that’s probably the most honest answer.
Humanity has never possessed this much technological capability.
We have nuclear energy.
Artificial intelligence.
Genetic engineering.
Global communication.
Advanced medicine.
Renewable energy.
Robotics.
Space technology.
We can manipulate systems on a planetary scale.
The problem is that our technological power is increasing faster than our collective wisdom.
We still have:
fear.
greed.
tribalism.
nationalism.
short-term thinking.
competition for status and power.
Those characteristics helped humans survive.
But when combined with technologies capable of changing the planet, they become considerably more dangerous.
The question I keep coming back to
The biggest question may not be:
Can humanity survive?
It may be:
Can humanity become wise enough to handle the power it has created?
Because we’ve already crossed a threshold.
We can alter the atmosphere.
We can destroy cities.
We can modify biology.
We can manipulate global information.
And we’re creating artificial intelligence that may eventually exceed humans in many forms of intellectual work.
The 21st century may therefore be less about whether humanity can become powerful enough to transform the world.
We’ve already done that.
It may be about whether our wisdom can catch up with our power.
Sources
The deeper research behind this conversation included:
Climate
- World Meteorological Organization — State of the Global Climate 2025
- IPCC — AR6 Synthesis Report
- UNEP — Emissions Gap Report 2025
- UNEP — Global Resources Outlook 2024
- UNEP — Adaptation Gap Report 2025
Energy and resources
- International Energy Agency — World Energy Investment 2025
- UNEP — Global Environment Outlook 7
Conflict and nuclear weapons
- Uppsala Conflict Data Program — 2025 conflict data
- Stockholm International Peace Research Institute — Trends in World Military Expenditure 2025
- Stockholm International Peace Research Institute — SIPRI Yearbook 2026
Artificial intelligence
- Stanford Institute for Human-Centered AI — AI Index Report 2026
- International AI Safety Report 2026
- AI Impacts — survey of 2,778 AI researchers
This is a record of a conversation with an AI, not a prediction of the future.
The future remains uncertain. The probability estimates above are subjective assessments, not scientific forecasts.
