Intelligence
The deep irony of our time.
First, I need to apologize for my absence. I am certain many (if not all) of you have deeply grieved my silence. Well, mend your torn garments. Un-gnash your teeth. I’m back.
In all seriousness, I have missed sharing thoughts through this Substack for the very reasons that have kept me offline: 1) so much is happening so quickly in our world and 2) I have undergone some personal hardship. On the latter, I’m hoping to share more in my next post, but suffice it to say these difficulties have hampered my creative impulses. More to come on that front.
The former reason that I have missed engaging with you all is what I want to devote this post to.
As I write this, mere days after the world inaugurated its first member of the trillionaire club, amid a deluge of seachange technological advancements on top of seachange technological advancements, and in the wake of impossibilities becoming commonplace, it feels important to slow the torrent, take stock of where we are as a global society, and foolishly prognosticate the future while also offering that brand of judgement you’ve come to expect (and appreciate?) from More Than Mundanity posts.
As with most of you, I regularly find myself in one of two conversations these days. The first starts something like this:
“All this AI, boy, I don’t know.”
The second starts like this:
“Where’ going to Jupiter with AI!”
I think both of those conversations tend to be a little myopic. They are probably more an amplification of a person’s pre-AI state of mind than anything else. If you are generally skeptical of technology, the undefined elite, the deep state, and most institutions, then you may find more affinity with the first conversant. If, on the other hand, you regularly pre-order the latest self-driving/propelling/operating whatever, lead board meetings, have a Tesla tattoo, and OK are also skeptical of most institutions, then you probably find yourself starting conversations like the second guy.
The problematic pro/anti AI dichotomy is really just a microcosm of so much in our society. It lacks nuance. It’s lazy. It’s just another topic for hot takes and judgement. Sure there are real concerns with AI (as linked above, Anthropic didn’t even release their Claude Mythos to the general public because their “AI models have reached a level where they can surpass all but the most skilled humans at finding and exploiting software vulnerabilities”). When, in a capitalistic society, a company pumps the breaks on its own industry-leading product, we may have tapped into something we shouldn’t have.
This gives me an opportunity to reclaim a quote that, while memorable in its movie adaptation is far better in the book. As another institutional skeptic once said, “[t]he Dwarves tell no tale; but even as mithril was the foundation of their wealth, so also it was their destruction: they delved too greedily and too deep.” The famous (italicized) part of this quote that Peter Jackson repurposed in the movie version of The Fellowship of the Ring is perhaps better for the part he cut. The quote continues in the book, “…they delved too greedily and too deep, and disturbed that from which they fled.” The misuse, abuse, and unintended destruction of AI is very real. AI very well may be the foundation of our wealth from which will flee. More on that below.
However, to toss AI out with the bathwater is both shortsighted and a little (read: a lot) naive. Like it or not, AI is here to stay, and not for purely bad reasons. AI is an extraordinary resource brimming with near-term and long-term potential. It can simplify onerous daily tasks. It can sort and quality-check our research (I’m talking about vacations, car shopping, and job searching, not academic research). It can make easy otherwise highly complex efforts like data analysis or deep dives into countries, candidates, and complex events in the news. It can revolutionize space exploration, discover cures that have alluded us longer than certain math problems, optimize city planning, and suggest real solutions to resource crises. AI is an invaluable resource that warrants exploration and (responsible) exploitation.
Those who are solely afraid of AI should be less cynical. Those who aren’t afraid of it all should be less naive.
The ubiquity of AI in its current state (unstatic as AI development is) should force us to use AI, dare I say it, more intelligently. What does that mean? I think there is a near-term answer to that question and a longer-term answer. The immediate reality of AI as harder to regulate but less dangerous. Whereas longer term, AI will be easier to regulate but far more dangerous. Easier (in theory) to regulate down the road in the sense that we will understand it better and thus better understand associated dangers: we are constantly becoming more familiar with AI taxonomy, functionality, and potential. We will also continue to regularly encounter ways in which it ‘misbehaves’ — (intentional em dash) from hallucinations to deception. Right now, we are squarely in the wild west of AI where everything seems possible, everyone has access to it, there aren’t many rules, and it’s still so novel we are overwhelmed with exciting known unknowns. So, in the near-term, we need to exhibit what people, collectively, have typically failed to do: being smart about assessing the merits of new inventions relative to their drawbacks. This is where sane, un-biased policymaking comes into play. Easy, right?
I think this has to be incremental. We should start with the lowest hanging fruit, that is to say, the areas we care to protect the most. Our kids. When trying to determine the balance between good and bad AI use, it’s helpful to think about how AI should be limited relative to what we want for our kids. Do we switch out math for ChatGTP subscriptions? Do we cancel art classes and give everyone unfettered access to DALL-E? Do we route the kids with really disturbing behavior to a bot that is trained on some undetermined, subjective set of empathetic traits? How about putting AI into school surveillance systems to track behavior, micro-expressions, even conversations to guide disciplinary action? I suggest these examples to demonstrate how seemingly well intended efforts can quickly spiral into unprecedented abuse (looking at you, institutional skeptics). I’m also hopeful considerations like these will force policymakers to draw clear red lines that put protections above possibility.
Given its increasing prevalence in our lives, leveraging AI without it supplanting so much that makes humanity uniquely beautiful (creativity, relational intimacy, artistic development, critical thought, moral sensibilities) is the challenge of our age.
Which leads to the longer term potential, and danger.
The world’s newest and only trillionaire, Elon Musk, recently gave an interview where he first discussed the future of AI, how data centers will be launched into space, how solar power in space will provide all the terawattage of energy we need for the exponentially consumptive AI compute demands (and possibly every other energy need), how Mars will be the back-up for when Earth destroys itself, and other wild considerations that are trademark Elon, prompting wide-eyed amazement and brow-furrowing concern.
After sharing his vision for the potential of AI - including the likelihood that, in the near future, only 1% of intelligence would be human intelligence - he was pushed to define a sort of mission statement about his grand vision of relatively unbridled AI progress that will ultimately build the robots that make the AI that build the robots that take us to Mars and beyond. He landed on “maximizing the scale and scope of consciousness.” The hosts of the show pushed him on this, especially after he remarked that, given there will be “a million times more silicon intelligence than there is biological, it would be foolish to assume there would be any way to maintain control over [AI].” In trying to qualify his mission statement, Musk kept using the word “truth.” It became clear that by “truth” Musk meant the verifiable, objective laws of physics. He noted that physics is the only definer of that which is real. Only physics can guide the unwieldy potential of exponentially intelligent and self-sufficient AI. Only the laws of physics can serve as the blueprint for what we should expect from and train AI to be. The laws of physics are true.
While they are true, they are not all of truth.
Musk’s attempt to answer the large, increasingly relevant question of our time that is inextricably tied to the longer term future of AI among humans was, well, logical. But I believe it is wildly insufficient. By narrowing truth solely to that which is mathematically verifiable forces us to toss out the reliability or even baseline merit of things like, human rights, compassion, forgiveness, reconciliation, beauty, and even love. Beyond their ability to contribute to a Darwinian take on the purpose of life (i.e., perpetuate the gene pool), these broadly valued virtues are incompatible with math-driven progress in the Muskian view of the future such that these virtues can, and maybe even should be discarded. Musk said that by programming AI to say things it doesn’t believe is programming AI to lie. And he’s not wrong. But this is the problem. If AI should be solely oriented to the physical dynamics of existence, and AI will ultimately run most (if not everything), then we are recoding our existential infrastructure to preclude any matter of the soul.
Taking a step back, I’m not saying AI=evil (read above). I am saying that a logical conclusion of AI’s unmitigated progression is a world in which the core of all our systems (AI) could make highly consequential decisions that preclude human conceptions of good in the name of actual objectivity. So unlike Musk, I think the answer is to ensure a virtuous element in AI model training (teach AI what is congruent and incongruent with our broader human understanding of right). But more importantly, I think it is incumbent upon us to choose responsible AI use and responsible AI disuse in the name of a more holistic truth. Unlike Musk, I do believe in an objective right that extends well beyond physics. I do believe God created all things including both physics and the very notion of good that has been corrupted, adopted, modified, domesticated, and exploited throughout all human history but that still echoes through the charitable and gracious moments in our world.
While the notion of good is extremely subjective in the global society rendering the infusion of good into AI futile at best, it doesn’t mean we shouldn’t still try to do it. Sure we will run up against incongruous definitions of good in training AI. This is already happening, which is why some people use Gemini or Claude or ChatGPT or Grok. The models are oriented on different notions of “right” responses elevating affirmation or objectivity or historicity. But protecting our kids, guarding against human exploitation, blocking violent and nefarious aims should serve as a baseline for both policy development and model fine-tuning.
In conclusion, I think people’s fear of AI is justifiable not because AI is inherently bad but because those of us who use it are. But while, as a Christian, I believe human nature is been deeply corrupted and necessitates holistic redemption, I also believe we are a remarkable creation overflowing with beautiful, astounding creative potential (e.g., writing The Brothers Karamzov, composing Don Giovani, and inventing AI). So let us value the virtuous on and offline. And when we lean into the winds of progress and the exciting potential of AI and whatever comes after AI, let’s not do so at the expense of protecting the sacred and prioritizing the righteous. Let’s be a little more nuanced and comprehensive with our understanding of intelligence.