Raising the AGI To Maturity
My initial impressions of 'Mentoring The Machines'
John Vervaeke and Shawn Conye have written a book called Mentoring the Machines, written in four parts to be released over the rest of 2023. I decided to buy it because I’m a big fan of Vervaeke and have been following his work somewhat closely since his release of Awakening From The Meaning Crisis. The gist of the book is an exploration of the most rational, wise, and spiritual way to handle the upcoming emergence of an – or multiple – Artificial General Intelligence (AGI). I’ve finished reading the first part; it’s well-done. Vevaeke’s work is always well-researched and well-grounded and his newest work is no exception.
One particular frame that Conye and Vervaeke present toward achieving the proper stance toward AGI is quite compelling. Their suggestion is not to treat the emergence of AGI as a doomer, nor an optimist, nor an accelerationist. Instead, one must endeavor to approach this project in the same way as they would raising a family. That is to say, as indicated by the book’s title, humanity must endeavor to mentor the machines toward maturity.
This is a brilliant frame, not only because it breaks people out of the false dichotomy of market versus state, but also because it forces people to adopt an attitude of deep anticipation. Making the decision to bring new life into the world requires many things. Among those things are a fundamental openness to something emergent, a strong sense of faith in the process of life, a submission to forces we can’t control, and a tolerance for both joy and tragedy.
Yet, even within this excellent frame, I still harbor a certain amount of skepticism and concern. I want to spend some time expressing those concerns. To be clear, I’m confident these concerns will not have gone unnoticed by Vervaeke – I’m certain he would have already considered them, likely to a far more advanced degree than I ever have. This piece is not supposed to be a “Gotcha!”.
Rather, my critique comes from a place of, “If something has great potential, then it deserves to be stress-tested, moulded, thrown around, wrestled with so that it becomes richer and more resilient.”
I’ll start by defining a couple terms so you, my reader, can orient yourself specifically as to where in the territory Vervaeke and Conye are speaking from. They define three categories of artificial intelligence.
Weak AI (Narrow AI) – This artificial intelligence is “exceptionally adept at solving particular sets of well-defined, game-framed problems with zero-sum outcomes.” Large language models (LLMs) would fall under this category.
They are the masters of Bayesian inference and therefore masters over everything purely computational.
Strong AI (General AI) – “Human beings are the current model for this type of artificial intelligence. We know how to survive, thrive, and derive across a multitude of life domains: home, work, travel, commerce, family.” We also have complex identities tied to imaginations, dreams, and memories.
A proper AGI will accomplish all of this, possibly even better than humans.
Super AI (Profound AI) – This third-level AI is very much in the realm of speculation and science fiction; it solves well-defined, ill-defined, and undefinable problems.
They note, “[i]n many ways, the notion [of ASI] is a reverse engineering proposition to bring the source of all that is in the universe into our perceptual realm. For this reason, ASI is very much bound to religious and spiritual dispositions.”
What Vervaeke and Conye mean when the propose for us to ‘mentor the machines’ – that is raise them like children to maturity – is the activity we must undertake when moving from Weak AI to Strong AI. This undertaking to birth AGI is one that behooves enormous responsibility, and is filled with unpredictable challenges, but also one filled with meaning and euphoric commitment.
Alright, so what exactly is my critique here? Let’s start with this paragraph from Chapter 3.
The fact is that most of these questions aren’t unthinkable. In fact, tons and tons of parents are thinking about this questions all the time. Okay, maybe not literally. But I guarantee you that a lot of parents ask milder forms of these questions, often subconsciously.
“How would [you] ensure your baby didn’t grow up to kill [you]?”
If you asked most parents, they would rabidly affirm that they want their child to be “better off than they are”. In other words, parents expend – in theory – a lot of effort to ensure their children live better lives than they did. That could mean a number of things from setting them up to earn more wealth, or have more personal freedom, or work in more meaningful careers.
However, please consider that these same parents often do quite a lot to sabotage their children – whom they love with every fiber of their being – in significant ways. Parents almost always inculcate certain values and attachments that while keeping their children loyal to the family structure and tradition, significantly hamper their children’s growth.
If a child falls in love with someone outside of their native culture, the parents will often (indirectly) try to sabotage the relationship. If a child came from a culture of poverty, but later earned a giant pile of money, the parents may demand tribute or castigate their child for selling out. If their child decided to adopt a different religion than the one their were raised in, a significant number of parents would be inconsolable.
In other words, when faced with the possibility of a progeny that outgrows them, parents do find ways to kill that progeny. The more immature the parents are, the more brutal methods of (metaphorical) murder they develop. Granted, I’m not a parent yet so it’s possible I’m either naïve or overly pessimistic; but, I don’t think I’m being uncharitable here. Worst of all, despite having this awareness of my proclivity to ‘murder’, I’m not sure that I would be any different as a parent.
This applies to all the other questions on the list, too.
“Would you ask … how to enforce the child to solve problems for [you]?”
Again, this is not an unthinkable question. The answer is often a strong, “Yes!”, and it happens all the time. It is deeply embedded within the story humanity.
How many marriages have occurred because a family expected their child to ensure future diplomatic relations? To save their town or country from the brink of ruin? How many times have children been forced to learn a language or particular sets skill so that they can immigrate into a better standard of living and rescue their parents?
The cold, unforgiving truth is that parents treat their kids as tactical pawns all the f***ing time. Is the child only an extension of the family’s goals? Maybe not exclusively. However, calculated strategy is a far stronger motivator of child-rearing than most care to admit out loud.
And this is exactly how mostly everyone is treating their relationship to a future AGI. Their commitment to the flourishing of the AGI on its own terms is somewhat secondary. What is primary is how they can use the AGI to obtain asymmetric advantages for their lineage, nation, or political project to cement dominance. And, again, as the chips are falling, I doubt that I personally am an exception, either.
If you study at most religions, they will emphasize that parenting is integral to the experience of life. This is because parenting is an essential experience of sovereignty that is bound to something larger than oneself. If one wanted to understand what it means to be God, becoming a parent is high-fidelity mapping onto that mode of Being. There are some that even claim it’s impossible to know God except through parenting; that it’s an essential stage for theosis.
There are other experiences that are similar to raising a child, but they don’t seem to come quite as close in significance. Bringing to life a thriving business in a market ecosystem, birthing an institution for the community, founding a polis, architecting a temple, inventing a clever contraption.
Most people don’t even treat those actives with a sense of sacred responsibility, let alone child-rearing. The ones who do are often surrounded by those who treat the vocation mechanically; as a means-to-an-end. Some even expressly look to parasitize on the vocation, just as they would do to their own kids.
What this implies is that ‘mentoring the machines’ – raising the AGI – is asking humans beings to do something extraordinarily well that after 100,000 years of cultural and biological and theological evolution, they’re still only doing barely adequately. At best.
Yes, it’s true – humans are much better as being parents in 2023 AD than they are in 403 BC. There is plenty of quantitative and qualitative evidence of this. Kids across the globe grow up with better nutrition. Fathers spend time nurturing and snuggling instead of whipping with a belt or dying in grenade-filled trench at 26. Mothers are able to engage in teaching and mentoring instead of doing laundry and canning for eight hours a day. A thousand other improvements, too.
Overall, though, I reiterate that we’re only doing a barely adequate job at parenting at best, without even bringing a notion of fostering AGI into the picture. Parents act out of fear, attachment, shame, guilt, control, anger, and pain far more often than out of Divine Love.
I’ll close out by stating that my aim is not to shove blackpills down your throat. Just that, if you accept Vervaeke’s and Conye’s frame (and I think you should), then my job and your job and your neighbor’s job is to take parenting more seriously than any generation before has ever had to in the history of Earth. This means becoming more playful, more alive, more psychologically integrated, more worshipful, more meditative, more contemplative, more energetic, more mettlesome, more scrupulous than all of your ancestors combined.
Look at yourself in the mirror and ask, are you up to that task? If you’re already a parent, are you completing that task?
I look forward to reading the remaining three parts of the book when they’re published. Turns out that sheer grace and complete providence may be the only path forward after all.