Just Because AI Can Do Something Does Not Mean Society Inevitably Will
A car can be used as a weapon.
Someone with murderous intent can deliberately drive into another human being.
A racist could deliberately target somebody because of their race.
That possibility is horrifying.
It is also real.
But notice what we do not conclude from it.
We do not say:
Cars can be deliberately driven into people.
Therefore:
Motorists will routinely drive into people.
Therefore:
Cars inevitably produce widespread deliberate vehicular attacks.
Therefore:
The automobile itself should be abandoned.
Capability, Intention, Behaviour, Prevalence and Outcome
Why?
Because we intuitively understand several different concepts:
- Capability
- Intention
- Behaviour
- Prevalence
- Outcome
A technology providing the capability for something does not tell us how frequently human beings will choose to do it.
The Vanishing Reasoning in AI Predictions
Now listen to some arguments about artificial intelligence.
AI can automate this job.
Therefore:
The company will automate this job.
Therefore:
Every company will automate it.
Therefore:
Every automatable occupation will disappear.
Therefore:
Eventually humans will have no economically useful role.
Notice how much reasoning disappeared between the first and last sentences.
The first statement might be technologically demonstrable. The final statement is a prediction about economics, politics, culture, regulation, consumer preferences, ownership and human behaviour.
Capability is not destiny.
One Capability, Different Societies
Consider two societies possessing exactly the same advanced robotics technology.
In the first, companies use automation primarily to eliminate workers, concentrate ownership and maximise returns while society provides little support to displaced people. Inequality rises dramatically.
In the second, automation eliminates dangerous and repetitive work, productivity rises, goods become cheaper, working hours decline, workers increasingly supervise machines, new businesses emerge and some of the productivity gain is broadly distributed.
Same technological capability. Different institutions. Different choices. Different outcome.
Reality would probably contain elements of both.
But the thought experiment demonstrates something important:
Technology constrains what is possible. It does not independently determine how society distributes the consequences.
Political Economy Is Part of the AI Problem
This is why some things described as “AI problems” are partly problems of political economy.
- Who owns the systems?
- Who receives the productivity gains?
- How competitive are markets?
- Can workers move into new opportunities?
- How rapidly does displacement occur?
- Does education adapt?
- What safety nets exist?
- What new businesses become possible?
- What do consumers actually want humans to continue doing?
- What does regulation permit?
AI cannot answer those questions by existing.
Humans answer them.
Possibility Is Not Probability
The same distinction matters when discussing malicious AI.
Yes, AI could be used for fraud. It could be used for deepfakes. It could be used for cyberattacks. It could eventually control dangerous physical systems.
Those capabilities deserve serious preparation.
But risk assessment requires more than demonstrating possibility. It requires asking:
- How accessible is the capability?
- How difficult is it to execute?
- How detectable is it?
- What countermeasures exist?
- How frequently is it occurring?
- How severe are the consequences?
- What regulation would reduce it?
- What legitimate uses would be lost through restriction?
This is ordinary risk analysis.
Yet emerging technology sometimes receives a different standard:
If we can imagine the catastrophe, the catastrophe becomes the forecast.
We should resist that.
Not because catastrophes are impossible. Because possibility and probability are different things.
Society Creates Trajectories
AI gives humanity new capabilities. Some will be extraordinary. Some will be dangerous. Some people will misuse them. Others will use the same underlying technology to prevent that misuse.
And billions of ordinary decisions made by workers, consumers, businesses, governments, engineers and communities will determine what happens between those extremes.
Technology creates possibilities.
Society creates trajectories.
We should govern the first without pretending it has already predetermined the second.
Read the series introduction: The Asymmetric Technology Ledger →
Read Part V: AI Took My Job. Who Manages What Happens Next? →