From Writing to Artificial Intelligence, Humanity Has Had This Conversation More Than Once
There is something strangely familiar about the conversation surrounding artificial intelligence.
AI will make people stop thinking. AI will destroy jobs. AI will corrupt education. AI will place dangerous capabilities in the wrong hands. AI will make people dependent on machines.
Some of these concerns deserve to be taken very seriously. But there is something we should remember before convincing ourselves that humanity has arrived at entirely unprecedented crossroads:
We have been afraid of technology before.
Writing and the Fear of External Memory
Long before artificial intelligence, there was anxiety about writing itself.
In Plato's Phaedrus, Socrates recounts an argument against writing that sounds surprisingly modern. Writing, the argument went, would weaken memory. People would depend upon external information rather than actually knowing things themselves. They might possess the appearance of wisdom without genuine understanding.
Replace “writing” with “AI” and parts of the argument could comfortably appear in a newspaper today.
And yet writing became one of the foundations of civilisation.
It allowed knowledge to survive the death of the person who possessed it. Medicine could accumulate observations. Governments could maintain laws and records. Mathematics could become progressively more sophisticated. Literature could travel across centuries.
Writing did change human memory. The concern was not entirely imaginary.
But humanity discovered something considerably more important:
We did not have to keep everything inside an individual human brain for human intelligence to advance.
Printing and the Distribution of Ideas
Then came printing.
Suddenly information that previously required painstaking copying could be reproduced at extraordinary scale.
And again came anxiety.
What happens when anybody can distribute ideas? What about false information? What about dangerous political or religious material? What happens to existing authorities when they can no longer control what people read?
Those were not foolish questions.
Printing really did make dangerous information easier to distribute. It also helped make mass education, scientific communication, newspapers and widespread literacy possible.
When Danger Produces Knowledge
Then came railways. Then cars. Then aircraft.
Each brought dangers that were sometimes spectacularly real. Early automobiles could kill pedestrians. Aircraft could fall out of the sky.
Humanity's eventual response to aviation is particularly instructive.
We did not say:
Aircraft can crash; therefore humanity should abandon flight.
We asked:
Aircraft can crash; what would make flying acceptably safe?
And around that question grew pilot training, maintenance regimes, air-traffic control, weather forecasting, accident investigation, redundant systems, aviation regulation, flight recorders and entire professions devoted to making aviation safer.
The danger did not merely produce restriction.
It produced knowledge.
Radio, Television and Institutional Change
Then came radio and television.
Some religious communities regarded television as morally dangerous. Parents worried about what children would see. Critics worried that people would stop reading, families would stop communicating and passive entertainment would damage society.
Decades later, the irony is difficult to miss.
Some religious organisations that once regarded television with suspicion became sophisticated users of television broadcasting. Today churches livestream services across continents.
The feared technology became part of the infrastructure of the institution that once feared it.
The Internet: An Accurate but Incomplete Risk Ledger
Then came the internet.
We were warned about pornography, fraud, strangers communicating with children, misinformation, hacking, social isolation and privacy.
Again, many of those concerns proved correct. The internet did not turn out to be harmless.
But imagine assessing the internet today using only that list.
We would have to ignore email, video calling, digital banking, online education, scientific collaboration, telemedicine, e-commerce, cloud computing, open-source software, search engines and near-instantaneous access to enormous quantities of human knowledge.
The risk ledger would be accurate.
It would simply be incomplete.
The AI Conversation
And now comes AI.
The temptation is to choose one of two camps.
AI will solve everything.
Or:
AI will destroy everything.
Neither is particularly useful.
History offers a better lesson. The critics of transformative technologies are often correct that a danger exists. Where we repeatedly struggle is in determining how much weight that danger deserves relative to benefits that have not yet been invented.
The Benefits We Could Not Yet See
A person watching an early automobile frighten a horse could see the problem immediately. They could not see an ambulance carrying a critically ill patient rapidly to a modern hospital.
A person worrying about television in 1955 could see a child spending too long staring at a screen. They could not see a university lecture broadcast globally.
A person encountering the internet in 1995 could see pornography and online fraud. They could not fully see cloud computing, remote working or a surgeon discussing imaging instantaneously with colleagues thousands of miles away.
And that creates a profound imbalance whenever a transformative technology arrives.
The harms can often be imagined immediately.
Many of the greatest benefits have not been invented yet.
Safeguards Without Abandonment
That does not mean we should ignore AI risk.
It means we should stop confusing the identification of risk with the completion of analysis.
Cars required seatbelts. Aircraft required air-traffic control. Medicine required clinical trials. The internet required cybersecurity.
AI will require its own safeguards, institutions, skills and boundaries.
Perhaps the mature question has never been:
“Can this technology cause harm?”
Almost every powerful technology can.
The better question is:
“Can we understand the risks, reduce them intelligently and still capture enough of the benefits to leave humanity better off?”
We Have Asked This Before
We have asked versions of that question before.
And fortunately for us, our ancestors did not always answer it with fear.
Read the series introduction: The Asymmetric Technology Ledger →