By Asha Geire (originally published under The Traveling Privacy Lawyer blog on Substack on June 12, 2026)
This week I’ve been thinking about complexity theory, artificial intelligence, and whether AI was less an invention than an inevitable consequence of humanity’s development.
Much of the public conversation about artificial intelligence seems to revolve around whether AI is good or bad.
The debate often resembles earlier arguments about the internet, industrialization, nuclear energy, or even the printing press. One side emphasizes transformative benefits. The other emphasizes catastrophic risks.
I don’t know if AI is inherently good or inherently bad, but I suspect it was probably inevitable.
I am not into technological determinism, but I have been thinking about this issue slightly differently lately. My thinking has been influenced by the work of physicist and complexity theorist Sara Walker, who studies how life evolves and how information becomes increasingly capable of organizing matter and influencing the world. While the details of her work are far more sophisticated than anything I can summarize here, one of the ideas I find compelling is that life appears to move toward increasingly complex forms of information processing.
Consider the arc of life on Earth. Single-celled organisms gave rise to multicellular organisms. Nervous systems emerged. Brains evolved. Humans developed language, allowing information to survive beyond individual lifetimes. Writing expanded that capacity further. Scientific institutions created mechanisms for accumulating knowledge across generations. Computers and the internet dramatically accelerated the storage, transmission, and application of information.
Viewed through this lens, artificial intelligence does not appear as a sudden anomaly. It looks to me more like the latest step in a long process through which information becomes increasingly capable of understanding and shaping the world around it.
That does not mean AI is destined to succeed, nor does it mean every development in AI is beneficial. It simply suggests that once human civilization reached a certain level of complexity, the creation of increasingly sophisticated cognitive tools was a likely outcome.
If that is true, then the AI debate should focus entirely on how these systems should be governed.
And on that front, history offers some perspective, though only up to a point.
The Industrial Revolution produced extraordinary gains in productivity and living standards. It also produced dangerous factories, polluted rivers, unsafe working conditions, and exploitative labor practices. Societies did not have all the answers at the outset. Instead, innovation often moved first and governance followed. Over time, we developed labor protections, environmental regulations, workplace safety standards, and other institutions that reduced many of the harms while preserving the benefits.
The Information Age followed a similar trajectory. The internet transformed communication, commerce, and access to knowledge long before societies fully understood its consequences. Only later did we begin confronting questions about privacy, cybersecurity, digital monopolies, misinformation, and the social effects of algorithmic systems.
In both cases, humanity had something important on its side: time.
The harms were often serious, but they were generally reversible or managable. Polluted lakes could eventually be cleaned. Unsafe workplaces could be reformed. Harmful business practices could be regulated. Societies could learn through trial and error, sometimes at significant cost, but without fundamentally losing control of the technologies themselves.
I am not convinced artificial intelligence affords us the same luxury.
If AI represents another step in the long evolution of increasingly powerful information-processing systems, then it may also represent a discontinuity from previous technological transitions. My concern is not simply that AI could create new harms. Every transformative technology has done that. My concern is that some harms associated with sufficiently capable AI systems may be difficult or impossible to reverse once they occur.
A highly capable model released broadly into the world cannot be “unreleased.” Dangerous capabilities, once widely distributed, cannot easily be recalled. Autonomous systems, if deployed recklessly, may produce consequences that emerge faster than institutions can respond. Even if many of the most extreme concerns ultimately prove overstated, the possibility of irreversible harm changes the governance equation.
For that reason, I am skeptical of the argument that society should simply allow AI to develop freely and address problems later. That approach may have been adequate for previous technological revolutions. It may not be adequate for this one.
This does not mean innovation should stop. Nor does it mean every new model requires heavy-handed regulation. It does suggest that baseline rules, safety standards, and governance frameworks should be developed alongside the technology rather than years after the fact.
One example of the tension around governance can be found in the debate over open-weight versus closed AI models. Open-weight models make their underlying model parameters available to the public, allowing researchers, companies, and individuals to run and modify them independently. Closed models remain under the control of the organizations that develop them and are typically accessed through APIs or hosted services.
Neither approach is obviously right or wrong.
Open-weight models can promote innovation, transparency, competition, and scientific research. They reduce dependence on a small number of powerful companies and allow broader participation in technological development.
At the same time, increasingly capable models may create new risks when they can be freely modified and deployed without meaningful oversight. Closed models allow developers to implement safeguards, monitor misuse, and respond to emerging threats, but they also concentrate power in the hands of a small number of organizations.
Reasonable people can disagree about where the balance should be struck. What matters is that this debate focuses on governance rather than on simplistic claims that AI is either salvation or doom.
In my opinion, the future will not be determined by whether AI exists. It will be determined by whether we are thoughtful enough to establish timely rules that allow us to benefit from it without losing control of it.