Beyond Intelligence No. 1
Artificial intelligence is becoming more deeply embedded in everyday life. As its role in culture, society, and financial markets grows, I expect to return to the subject often. I plan to use this space to think through this evolution as it unfolds. This article is the first installment in a series of thought pieces on the subject.
Spoiler Warning! Skip ahead if you want to avoid spoilers for Silo, seasons 2 and 3.
My wife and I recently finished watching the third season of the Apple TV series Silo, based on Hugh Howey's novels written between 2011 and 2013. For unfamiliar readers, the basic premise is that there is a group of people living in a vast underground bunker, a silo, and they have lost much of their understanding of the outside world. What is interesting about the story for today’s discussion happens in the second season. The viewer finds out that one of the leaders of the silo is in contact with a disembodied voice that appears to be an all-knowing artificial intelligence.
Today, this seems like a completely reasonable inclusion in a sci-fi series set in a dystopian future. However, in a pivotal scene, a character argues that the voice cannot be a computer because it showed compassion towards the leader when they faced a difficult decision. Surely, no machine could do such a thing!
That was a perfectly plausible line of reasoning when the story was written. Today it sounds almost quaint.
Spoiler Warning is over.
Most of us interact with AI systems regularly. Whether it is ChatGPT, Gemini, Copilot, or even a simple Google search. Computers now routinely communicate in a way that can feel thoughtful, empathetic, and reassuring. Whether the machine truly understands what it is saying is a deeper question. The practical reality is that many interactions already feel distinctly human.
What was pure science fiction only fifteen years ago has become part of everyday life. For some, this is more of a pressing issue than for others.
In academic settings, the conversation is particularly relevant. Student performance has struggled to recover from the disruption caused by the pandemic (OECD Data), and now students must contend with having access to increasingly capable artificial intelligence. “Back in my day,” we complained about teachers restricting the use of calculators and computers during exams. We reasoned that those tools were already widely available in the real world, so we should be able to use them.
A similar argument is being made by students today. But the real question is no longer whether students should have access to useful tools. The question is what skills remain uniquely valuable when the tool can not only calculate an answer but also explain it, defend it, and present it in polished prose.
The issue becomes harder to ignore with every new headline that announces the cutting-edge frontier model has solved a question that has perplexed human scientists or mathematicians for years. Recently that was the headline from OpenAI, who announced a proposed solution to the Navier-Stokes existence and smoothness problem, one of the famous Millennium Prize Problems in mathematics (OpenAI Announcement). Whether the proof ultimately withstands scrutiny is for mathematicians to determine. The more important point for the rest of us is that a system built by an AI company generated work sophisticated enough that experts are taking it seriously.
Tristan Buckmaster, a prominent researcher in the field, was quoted, “This is a Deep Blue–Kasparov moment,” referring to IBM’s Deep Blue computer that beat renowned chess champion Garry Kasparov in 1997 (Scientific American).
Some readers will remember how significant that event felt at the time. Today, no serious chess player expects to defeat a chess engine set to the highest difficulty. Your smartphone would likely overwhelm Magnus Carlsen. What once seemed like a historic breakthrough has become mundane.
Could we expect something similar to happen with today’s AI? Deep Blue occupied two server racks and could perform about eleven billion floating-point operations per second. The computational part of a modern AI system that would be capable of solving the Navier-Stokes problem would be built from roughly fourteen Nvidia Grace Blackwell server racks. That type of system can perform, in the order of, twenty quintillion operations per second. The precise comparison is imperfect because the machines were designed for very different purposes, but the magnitude of the difference is difficult to ignore. The modern system possesses roughly a billion times more computational throughput than Deep Blue.
Can we ever expect this type of computer power to sit in the palm of our hand? Probably not. There are physical limitations to semiconductor density that we are already running up against. That being said, we do not know how quantum computing systems might evolve over time and, when integrated with traditional binary computers, they may be capable of similar feats.
The rate of technological progress is astounding. More importantly, the scope and scale are completely unobservable and incomprehensible to humans. We can attempt to wrap our minds around what it means to control enough computing power to solve a 100-year-old math problem, but can we ever truly? For example, twenty quintillion (the number of operations that fourteen Grace Blackwell racks could perform every second) is around seven million times the estimated number of trees on Earth (an estimate for number of trees on Earth is three trillion).
While this is a completely unhinged analogy, I think that is where we are. As humans, we are simply not built to fully grasp the extraordinary scale of such numbers or to have the ability to accurately predict the direction and speed of technological progress. And do we need to? When we have systems that can hold such precise data in memory, turn them over, produce real results, and then communicate the results with us in a conversational tone. It behooves us to acknowledge our limitations and get on with what we are good at. Connection, creativity, and courage. Though our current version of AI can mimic these traits, I do believe this is what sets us apart.
Disclaimer
This material is provided for informational and educational purposes only and is not intended as, and should not be construed as, investment, legal, tax, or accounting advice. The information is general in nature, does not take into account any individual’s objectives, financial situation, or needs, and should not be relied upon as a recommendation to buy, sell, or hold any security or to engage in any particular investment strategy. Nothing herein constitutes an offer to sell or a solicitation of an offer to buy any security or investment advisory services, nor is it intended to create an advisory relationship. All investing involves risk, including the possible loss of principal. Past performance is not indicative of future results, and no assurance can be given that any strategy will achieve its objectives. Any forward-looking statements, opinions, or estimates are as of the date indicated and may change without notice. Information has been obtained from sources believed to be reliable; however, accuracy and completeness are not guaranteed. If index, benchmark, or third-party information is referenced, it is provided for illustrative purposes only. You should consult with a qualified professional regarding your specific circumstances. Additional information about the adviser, including Form ADV, is available upon request.