Introducing an essay in 3 parts
With the immediate landscape of AI shifting so quickly, it can be hard to keep the broader view in sight.
In just the last few months, OpenAI published findings that, during safety testing, its o1 model would sometimes strive to deactivate its oversight, deny doing so when asked, and seek to exfiltrate its “weights” or distinctive neural architecture parameters so that it could not be shut down. Meta’s and Alibaba’s open-source models have proven capable of creating functional copies of themselves. OpenAI rolled out access to its AI agent “Operator,” and let third-party developers use it to build applications that can autonomously browse the web and write code. DeepSeek demonstrated that very impressive AI “reasoning” can be done at a fraction of the cost previously thought possible, and then Grok 3 showed that a full-steam-ahead approach of building data centers continues to improve the state of the art.
Will AI Be Alive?
An essay in three parts
Introduction
1. Gaining Situational Awareness About the Coming Artificial General Intelligence
2. It Will Seem to Be Alive
3. We Must Steward, Not Subjugate Nor Worship It
But we also have a helpful preview of what’s to come. “Situational Awareness: The Decade Ahead,” a June 2024 essay series by former OpenAI programmer Leopold Aschenbrenner, makes a strong case that artificial general intelligence is likely to be here within two years. “These machines will outpace college graduates,” he writes. “By the end of the decade, they will be smarter than you or I; we will have superintelligence, in the true sense of the word.”
This following essay proceeds in three parts, borrowing the “see, judge, act” framework developed by Cardinal Joseph Cardijn, a Belgian priest and activist on behalf of industrial workers. To act rightly, we must judge prudently. And to judge prudently, we must see clearly. The stakes are profound: understanding the fate of humanity while our leading technologists expressly seek to build what many call a successor species to the human race.
So, first up is seeing clearly, with Aschenbrenner as our guide. Where are we, and where does the path ahead lead? The answer is: to artificial general intelligence and robots with the ability to achieve goals in a wide range of applications.
Next is judging prudently, with help from Aristotle and his heirs. What would those embodied intelligences be? The disconcerting answer is: in a partial but real sense, an artificial form of life.
Last is acting rightly. How ought we to respond to such a being? The answer is: neither as its master nor as its slave, but as its steward.
No one can be sure of the future, but we would do well to consider the implications of Aschenbrenner’s prophecy, if it’s true. There was a crux point in the history of philosophy and law when modern thinkers began arguing etsi deus non daretur — that is, making arguments that would be true even if God were not a given. While taking the time to consider reasonable objections, let us proceed ac si intelligentia artificialis daretur — as if artificial intelligence were a given.
If artificial general intelligence came to pass as Aschenbrenner foresees, then it would have a kind of life of its own, acting autonomously in the sense of having agency and perhaps even in the sense of following its own laws. And if this AI seems to live, then it will be not merely used or befriended, but by many worshipped. And so we turn to ponder this future, that perhaps we may avert the worst of it.
