Jurassic Park and the Folly of Artificial Intelligence
Don
Allen and Jessica Allen
Jurassic
Park is ostensibly
a story about resurrecting dinosaurs, but beneath the adventure lies a more
profound warning about technological progress. Its central problem is not that
the scientists fail to accomplish what they intended. They succeed
spectacularly. They bring dinosaurs back to life. They create an extraordinary
park. They demonstrate scientific capabilities that once seemed impossible. The
trouble begins because the future consequences of that success are
substantially unknown. In this respect, Jurassic Park provides a
compelling metaphor for the advent of artificial intelligence.
John
Hammond asks, in effect, whether humanity can create something extraordinary.
The more important question, however, is what happens after it has been
created. Artificial intelligence presents a similar dilemma. We can build
increasingly capable systems that write, reason, program, analyze, discover
patterns, and potentially contribute to scientific research. Yet knowing that
we can build such systems tells us very little about how their increasing
capabilities will interact with individuals, institutions, economies,
governments, and civilization itself.
The
deepest lesson of Jurassic Park may therefore concern unknown unknowns.
Hammond can anticipate familiar dangers: dinosaurs may escape, fences may fail,
computers may malfunction, and visitors may be injured. What he cannot
adequately anticipate is the behavior of a complex adaptive system once its
components begin interacting in ways beyond the designers' expectations. The
dinosaurs do not simply remain the objects Hammond created. They become
participants in an evolving ecosystem.
AI
may present an analogous problem. We often think of artificial intelligence as
a collection of tools: assistants, chatbots, programmers, researchers, or
automated agents. Yet as these systems become more capable and interconnected,
the important object of study may no longer be the individual AI. It may be the
larger technological ecosystem produced by millions of interactions among AI
systems and human institutions.
This
raises another lesson from Jurassic Park: control can be an illusion.
Hammond constructs fences, security systems, procedures, and safeguards,
believing that the system can therefore be contained. But complex systems
frequently discover pathways their designers did not anticipate. The famous principle
that "life finds a way" can be translated into a technological form:
capability may find a way.
The
greatest folly, then, may not be creating AI. Nor is it necessarily believing
that AI will become dangerous. It is assuming that the future version of AI
will remain within the categories we use to describe it today. Hammond thought
he was building a theme park. He was actually creating an ecosystem.
Perhaps
that is the most unsettling parallel. We are not necessarily building the
future we imagine. We may be building something whose consequences will only
become visible after it exists. The future, like Jurassic Park, has no fence
strong enough to contain everything we have not yet imagined.
Comments
Post a Comment
Please Comment.