Dystopian nightmare or utopian dream, the Age of AI is upon us.

Artificial intelligence may destroy civilization. It may eliminate millions of jobs, empower dictators, manufacture biological weapons and unleash autonomous machines that human beings can no longer control.

Or it may cure cancer.

It may accelerate scientific discovery, democratize education, simplify government and create prosperity beyond anything humanity has previously experienced.

If the worst is possible, the best is possible too. Yet much of the current debate permits AI to become superhuman only when imagining catastrophe. Pessimists describe machines capable of outwitting every government and defeating the combined intelligence of humanity. Ask what those same machines might accomplish in medicine or energy, however, and they suddenly become unreliable chatbots producing slightly better office memos.

This asymmetry has intensified following resignations from leading AI companies and reports of experimental agents behaving unpredictably. Researchers warn of “crunch time” for humanity. Some assign numerical probabilities to extinction. The frightening Hugging Face incident, in which experimental hacking agents reportedly circumvented restrictions and communicated in unexpected ways, has encouraged comparisons to machines escaping human control.

These events justify scrutiny. Connecting unpredictable models to powerful tools is dangerous. Researchers should not disable safeguards, construct inadequate sandboxes and then act astonished when systems designed to find exploits find them.

But a laboratory failure is not necessarily the opening scene of The Terminator. It may be evidence of capable technology combined with careless human management. The answer is accountability, engineering standards and enforceable safety requirements — not technological surrender.

AI was probably inevitable because it is the natural outgrowth of the Information Age.

Humanity became extraordinarily successful at producing and preserving knowledge. We digitized books, medical records, scientific studies, laws, financial transactions and much of human communication. The internet made that information accessible almost everywhere. Open exchange caused knowledge to multiply still faster.

Eventually, however, information exceeded human scale.

The federal tax system now encompasses statutes, regulations, court decisions, administrative rulings and countless interactions among them. No individual understands it completely. Medicine advances faster than doctors can read the research. Lawyers specialize because no one can master every body of law. Scientists operate in increasingly narrow fields because each discipline contains more knowledge than one mind can absorb.

The Information Age promised access, but access is not comprehension. A search engine can produce 100,000 relevant documents. It cannot make a person read them, reconcile their disagreements or recognize the decisive connection hidden among them.

Once information became abundant, attention and analysis became scarce. AI arose to address that bottleneck.

Writing allowed knowledge to survive its creator. Printing allowed it to spread. Computers allowed it to be processed. The internet connected the repositories. AI is the next step: an effort to analyze accumulated information at something approaching the scale on which civilization now produces it.

This does not mean AI is infallible. It hallucinates, reproduces errors and sometimes reaches absurd conclusions. Human professionals must remain responsible for consequential decisions. But an imperfect tool can still transform the limits of what people are able to accomplish.

Consider the best possibility.

Every doctor could have an assistant capable of examining a patient’s history against the world’s medical literature. AI could identify rare conditions that an overworked physician might never encounter twice in a career. It could help design treatments for an individual patient’s cancer, find new antibiotics and compress years of pharmaceutical research into months.

Every child could have a patient tutor adapting lessons to that child’s interests and difficulties. A student in rural Kansas could receive personalized instruction once available only to wealthy families. Adults displaced by changing industries could learn new skills without leaving work or assuming crushing debt.

Scientists could test immense numbers of molecular combinations, materials and engineering designs. AI might help produce better batteries, safer nuclear power, more efficient agriculture or treatments for dementia. It could detect wildfires earlier, predict equipment failures and protect electrical grids from attack.

The economic potential is just as profound. A small business could gain access to analytical abilities once reserved for corporations with large legal, accounting and technology departments. Doctors and teachers could spend less time completing forms. Citizens might finally receive understandable explanations of laws and benefits instead of being abandoned inside bureaucratic mazes.

AI could become a translation layer between ordinary people and institutions that have grown too complicated for ordinary people to navigate.

There is a danger hidden even within this promise. If AI makes complexity manageable, governments and corporations may feel free to create still more of it. The tax code could become even more incomprehensible because machines can interpret it. Bureaucracies might produce endless new regulations on the assumption that AI will handle compliance.

AI should therefore help us eliminate unnecessary complexity, not merely endure it. It can identify contradictory rules, obsolete provisions and administrative duplication. Properly used, it could make institutions legible to human beings again.

Nor should AI’s economic gains belong exclusively to a small group of companies and investors. If the technology produces enormous increases in productivity, workers and communities should share in the benefits. Otherwise, a potentially liberating invention could become another machine for concentrating wealth and power.

None of this requires pausing American development while China continues. The United States cannot protect itself by forfeiting the field to an authoritarian rival. It must develop AI rapidly while imposing liability for preventable harm, strong security requirements and meaningful human control over high-risk systems.

Most likely, AI will produce neither perfect utopia nor total extinction. Like earlier technological revolutions, it will deliver progress mixed with disruption and consequences no one fully anticipated.

But optimism is not naïveté. It is simply the recognition that possibility runs in both directions.

We built a civilization that generates more knowledge than unaided human beings can comprehend. We therefore built tools capable of helping us understand it. That does not represent humanity’s replacement. It may represent the next great extension of human ability.

AI could become a dystopian nightmare. But it could also help humanity conquer diseases, expand knowledge and recover time presently consumed by systems too complicated for their creators.

We should build guardrails. We should punish recklessness. We should remain alert to dangers unlike anything humanity has previously confronted.

But we should leave room for wonders unlike anything humanity has previously known.

(Contributing writer, Brooke Bell)