When I was doing my PhD, I was getting paid a $1400/month stipend as a graduate student and teaching assistant. It took me a total of 5 years to learn, review all literature, do the analysis work and write the thesis. Not all efforts had equal weight — there was much menial work and unproductive periods, and the actual research breakthroughs happened in a few weeks during serendipitous moments, mostly after conferences and brainstorming with others.
Current AI is still far from doing research tasks autonomously. But from the pure intellectual horsepower perspective, a model like O1/O3 can help with those tasks immensely.
I should note that a graduate student is way more than just the pure problem-solving brain. However, that inspirational spark, the core ideas that made me able to push the boundaries of science, was a very small fraction of the volume of entire work needed. If I had an O1 model today, I would maybe be able to finish my work much faster, or go way further. But fundamentally, constructing a researcher-assistant agent with an O3 model is just some more plumbing steps away. Now a senior researcher can have an army of research assistants to help them code, investigate, conduct experiments and analyze results.
The frontier of AI now is really at this level of sophistication. I feel we currently have a very strong component of AGI, mainly the reasoning engine. The rest is the externalities of connecting it to the outside world, to build this agent so it can capture the context and memory it needs and to interact with the outside.
At first glance, this story feels like a sign of humanity's progress and advancement. But at the same time, it gives me a sad feeling. To see that intelligence is so cheap and so accessible, it should be an expediting force in society for good. The AGI moment should be considered a groundbreaking moment. But we will be surprised to know it came and went without much notice, while society is still doing what it used to.
This is ironic that we have a very capable model that can solve the hardest math Olympiads and graduate-level problems but people really don't know what to do with this. I feel it will take a long time for abundant intelligence to diffuse and make a meaningful change in our civilization.
If you think about it, this is not actually that surprising. Even our society doesn't know what to do with its brightest minds, the same PhD graduates with biological brains! Some of the best physics and math geniuses go do algorithmic trading. Many talented individuals never get a chance to pursue their true potential. Rarely does any smart person find a job that is as intellectually demanding and that pushes their intelligence to its limits.
It is becoming apparent that the structure and fabric of our society doesn't need this level of intelligence after all. For example, look at the political system, how are the most important and consequential decisions made? Not so different from the Middle Ages when a king was ordering people on a whim.
Look at the enterprises, the healthcare system, the financial system… We barely see intelligence as the main driver of leadership.
The problem with the adoption of intelligence is not simply inertia but is the active resistance to a more intellectual way to do things. The aversion is rooted in the processes and structures that we have adopted to run our society. We have constructed it not based on our high level of intelligence but on the basis of power structures mostly borrowed from our ancestors both anthropologically and phylogenetically.
This lack of intelligent foundation in our society becomes less surprising when considering its development over the course of history; of course, intelligence is just one of the drivers of survival in ancient societies… But we do not need to remain trapped.
A Mental Experiment
I would like to suggest an exploratory question, a mental experiment: think how we can alter the path of our society to actually utilize this level of intelligence that is readily and cheaply available.
Let's make an analogy between a civilization's use of energy and its progress. Back 100,000 years ago, a large human society could be one with a sprawling temporary residence in a savannah with multiple groups sitting around campfires and preparing food. Maybe a group of a few hundreds with tens of campfires. They are using around 5kW energy per fire and say 1MW for the whole society. As our societies evolved, we now use 19 TW, 7 orders of magnitude more. See Kardashev scale for excellent detail.
If we construct a parallel for the use of intelligence, probably we have seen a much smaller progress. In prehistoric societies, an individual on average may have used their brain 1% of the time for discovery or solving a hard intellectual problem (say finding the best material for the red paint).
A thousand years ago, we started having dedicated individuals and institutions for intellectual work like early hospitals and universities. Today many jobs and functions require solid intellectual endeavor. (Ah, I am excluding writing those corporate emails as intellectual activity). Maybe today we utilize 10–20% of all biological intelligence available. So, assuming prehistoric population of 1M, we have increased our species' intellectual activity by only 4 orders of magnitude, much less than the 7 orders of magnitude we have seen in energy use. Well, now this is our moment to catch up!
Intelligence with Empathy
Before we start talking about using more intelligence, let's consider an important point. What do we use the intelligence for? Who is the beneficiary of its use and its exponential growth? To answer this question, we need to consider an equally important part of our human nature, our heart; our empathy.
We use 7 orders of magnitude more energy than prehistory but this energy use has not been equally and fairly distributed, not only its use but also the devastating impact it has had so far on the environment. So, keep this in mind before we think how to expedite our use of intelligence while maintaining our empathy towards all human beings, and broader life on earth. I actually believe an extensive use of intelligence in society is actually a driver or at least a contributor to a fairer world.
Transforming Society
So how can we alter the path of our society to actually utilize abundant intelligence and to make it a better one as well?
The first target is to expedite our transition to a post-work society, where abundant intelligence performs most aspects of production and service activities so that human intellect is free to dedicate itself to discovery and progress and be aided with AI tools to go even further. In this society, all people are free to utilize their intelligence to its fullest extent. Fairness is important so that we make sure we give this opportunity to all to fully realize their intellectual potential.
But most importantly, we need to transform the underlying structures of governance and the methods by which decisions are made. I'm not only referring to political institutions, but to the very core of how society is organized. Consider how information is gathered from every individual, how it's processed, and then transformed into decisions across different social strata. Observe, too, how disinformation persists in our current systems, and how choices are often driven by untruths, reactionary thinking, and populist impulses. This is alarmingly reminiscent of how past societies operated — predominantly through gossip, rumor, and myth. Historically, the channels of decision-making were under the tight control of a select few, ultimately catering to the ruling elite. Strengthening democracy, and grounding its decision making process in factual evidence and analytical reasoning, is therefore essential to enacting this fundamental overhaul.
Now, you can see the role of abundant intelligence in this paradigm, imagine AI fact-checking everything we see and hear. Imagine simulated digital twins of the world assisting us to foresee economic and environmental consequences of any decisions and policy. Consider AI assistants that can show biases in our beliefs and help us consider various points of view privately and in personal ways (very different from another human arguing with you). These "products" are not yet built but early research is showing promising results in the fact that AI can assist us in this direction.
I hope AI will usher in a shining era of human civilization where we become more rational and importantly more empathetic. I hope this pure and powerful intellect allows us to push the utilization of our collective biological intellect to even higher levels. The only type of intellect that is influenced by our hearts.
Notes
- I will write more about the interplay of intelligence and fairness in a separate post. This may not appear clear but I argue the reason our democracy in contrast to Roman Republic does not feed slaves to lions for entertainment is because of many intelligent people who changed our views over the two millennia.
- See literature on information-based governance theories.