What if an alien civilization, or AI we created, knows the secret of the universe but we can’t understand it?

In my religious-believing days, Eastern religion variety, I loved the term, ultimate reality. I used it in talks I gave to members of the religious organization. I used it in books that I wrote which were based on the spiritual teachings I once embraced.

My assumption was that whatever ultimate reality was — my guess was All-Pervading Conscious Energy — the problem lay in learning its nature, not in understanding what that nature was, once it had been revealed.

Meaning, we humans had the ability to grasp what ultimate reality is all about. While it wasn’t an open book, the pages of ultimate reality could be read and understood if the book were to be found.

Well, maybe I was wrong. In fact, I now think there’s a very good chance I was wrong.

Today I finished one of the concluding chapters in Do Aliens Speak Physics? by Daniel Whiteson and Andy Warner. Whiteson is a particle physicist. In “How Smart Are the Aliens” he and Warner start off with a thought experiment.

Imagine “alien visitors had offered to share the secrets of the universe with exactly one human representative.” After an exhaustive worldwide search, “Anika had aced every exam, beating out older, much more established names to win the chance to be the first human student — the only human, the only student — welcome at the Alien Physics Academy.”

Anika is intimately conversant with AI, artificial intelligence. She takes her AI along with her to the Alien Physics Academy. At first she is able to follow what is presented, including the complex mathematics, with no difficulty. Then the lessons get more difficult. She needs the AI to help her grasp certain concepts. And then…

The theory took a dramatic twist, upending the story and heading in multiple new directions at once. Instead of building toward a natural simplicity, the story of the Universe branched, and branched again, like a fractal. Anika could no longer follow the big picture. It was like the airplane had come apart and she was in midair, a falling dream you can’t wake up from.

…Soon the alien cube’s concepts were a blur. Anika had lost her mental footing long before and was relegated to nodding along at the onslaught of ideas, taking what notes she could to help her future self untangle everything. It was like being swept along in a wave. Moving despite herself, Anika allowed herself to marvel at the fragments she could understand and love what she did not. The unknown, the divine.

…In the days that followed, Anika and the rest of humanity dug into the notes that she and her AI had taken. The majority was an incomprehensible jumble, like something a freshman might scribble down in a master’s course if they thought the professor was staring at them.

…Anika tried to work with her AI like she always had, but found it frustrating in unfamiliar ways. She was able to follow the story up to a point, beyond which she just couldn’t hold it in her head. Thew were too many pieces, all in motion simultaneously, for her to make an effective mental picture. The AI insisted that the math checked out, and other independent AIs were asked to verify it.

The cube floating in low Earth orbit stayed inscrutably quiet, so humanity figured they just needed to trust the AIs until they got the hang of it. New fields of research blossomed on the backs of these alien-math-powered AIs. All were beyond the power of their biological masters.

A couple of decades passed before the cubes suddenly materialized again. They were ready to bestow another gift. But this time, the cubes invited the notebook. Anika and the rest of humanity were told to stay home.

In both mysticism and everyday experience, there is more to reality than can be expressed. We always know more than can be put into words, numbers, pictures, or other symbols. Our love for someone can never be fully captured in any description. All we can do is try to express the inexpressible as best we can.

But this is different from the alien thought experiment I shared above. Neither Anika nor any other human is able to know the secrets of the universe being conveyed by the aliens. Later in the chapter, the authors write: “It’s possible there is an answer,  but one we can’t understand. It might be, as Chomsky says, ‘a case where cognitive capacities just don’t happen to conform to the nature of the universe.'”

We’re getting glimpses of this in the increasing concern that AI models are becoming too intelligent too fast. We humans are at risk of losing control of what we’ve created. Here’s a gift link to a September 13 New York Times story, “What Anthropic’s C.E.O. Argued in His Call for Slower A.I. Development.” The story had a link to the 3,800 word letter from Dario Amodei, “We Must Pace the Frontier.” Amodei said:

My first concern is that, since roughly this summer, AI has been advancing drastically faster, driven primarily by AI’s growing ability to build the next generation of AI. This dynamic is called recursive self-improvement, and it is starting to happen across the industry, including at Anthropic, as we and others have described. Left unchecked, it could outrun our ability to understand and control these systems, and so must be pursued very carefully, if at all.

…Despite all the progress, we still only understand a tiny fraction of what goes on inside these models. A focused effort to improve our interpretability techniques, even faster than we currently are, could make profound progress in 1–2 years, and would have ample experimental material based on the incidents that have already occurred.

So there’s a valid concern that just as we humans may not be able to understand what an alien civilization wants to teach us about the nature of reality, we also are becoming increasingly unable to understand the inner workings of AIs. An AI may give us accurate knowledge about reality, but there’s a risk that we won’t be able to understand what is being conveyed to us.

Kind of like how our dog is fully able to eat and enjoy the food my wife and I put in her bowl, yet is clueless about how to obtain that food or what it is made of. Whiteson and Warner say:

We do have some experience  trying to absorb radically new ideas, like quantum mechanics, but so many decades after its discovery, our continued failure to completely wrap our brains around it might be a sign that some ideas lay beyond our abilities. We may be able to extend our mental capacity with artificial intelligence, but we still risk getting answers without explanations. Before we can teach our AI to talk to aliens, we need to first teach our new AI assistants to speak our own mental language.