internetguy.dev/posts/on-riemann-navier-stokes-and-the-first-principles-of-living-a-fulfilling-life
all posts

on riemann, navier stokes and the first principles of living a fulfilling life

I’m sure many of you have heard of the recent breakthroughs on the (generalized) Riemann Hypothesis (GRH) and the unforced Navier-Stokes.

If you haven’t, some links include:

I will preface this by saying I am no expert in math or computer science or any of these fields these problems are situated in. In fact, I’m probably pretty average at math/cs. I am a just a freshman in college after all! Feel free to e-mail me or contact me on Discord if you disagree (or agree) with what I say!

Yet I think these problems are cool and obviously worth talking about. I’ve also been thinking a lot lately on more fundamental aspects of my life in this new “AI revolution” and how I can continue to build cool things, work on interesting and fun problems, and generally enjoy my life.

Over the past few months I have been hearing a lot more doomerism about AI and how it will take all our jobs, kill any & all intellectual work, etc etc. I really disagree with all this hopelessness. There are so still so many unsolved problems with AI and in all sorts of fields of research, we certainly have not “solved” humanity or anything like that. As we continue to solve problems I suspect even greater problems will come to mind and we will continue to complete greater advancements with the help of AI as a tool.

I’ve always seen myself as quite an optimist with a very pleasant expectation for the future. I’d like to separate this blog post into a few sections; first I will address my thoughts on the recent advancements in mathematics before talking more broadly on the tenets I will strive to live by going forward.

Regarding GRH and NS and math

I have many friends studying pure mathematics that are in shock and even a little depressed at the (very probably) convincing proof of the generalized Riemann Hypothesis. Many people say that pure math is “dead” or mathematics is “solved”.

Looking ahead in the short/medium term, mathematicians will still be in demand and will have a place in society. As Alpoge himself puts it,

I intend to do my best to understand and impart understanding of this argument and its certain simplifications, to speak about it, to write it properly.

Aside from technically impressive breakthroughs, the greatest mathematicians have always been excellent communicators of their results. Alpoge is a great mathematician, no doubt, and he is keenly aware that his true value lies in his ability to speak and write about it properly. To be a good communicator for humans, a mathematician must intuit and comprehend knowledge like humans. The best way to do this is to be a human. LLMs (at least in their current form) will never be great communicators to humans because nothing in their training or architecture allows them to understand the common mistakes that humans make when learning and the various analogies that the human brain makes while building some mental model of knowledge.

Some of the most educational conversations I’ve had with others begins with “you know how if you…” or “remember when you learned…” Such a phrase is a beautifully human thing because only a human teacher would understand the common pitfalls of learning a topic and exactly how to steer a student out of a pitfall or provide an accurate analogy for the student to build their intuition. No LLM would truly be able to recreate the connection of making a mistake and getting un-stuck from a problem because no LLM has never made such a mistake.

All this is to say we still need competent mathematicians that have great intuition for problems and the ability to communicate their ideas and solutions. Such a skill remains valuable and will remain valuable for quite some time in the future.

In the long term, I believe in the beauty of problems and the ability of us humans to find greater and greater problems to solve. There are many conjectures and hypotheses out there still waiting to be proved, formalized, and understood. Some of them problems (like the Collatz Conjecture) are beautifully simple on the surface but have pushed us to make breakthroughs in every other area of math. While the Riemann Hypothesis stood as a “final boss” of math for decades, I am optimistic we will find another, greater “final boss” to turn out attention to soon. I doubt we will ever run out of problems to solve.

History tells many tales of great “final boss” problems being trivialized by technology. For centuries humans worked on the problem of efficiently copying documents. We developed contraptions and printing presses and stencils and carbon paper all for these efforts to be replaced by a screen and electrons. Revolutionary new technology will always conquer many old problems but will always provide new questions to answer. I am firmly and deeply optimistic that centuries and millennia into the future we will have interesting problems to solve.

Regarding taste: Will AI be able to solve all these problems by itself?

This is a valid question and something I’ve been thinking about as well. “Taste” is one of the hottest words right now in Silicon Valley. Broadly, it means being able to spot and define problems that are “worth” solving and being able to build solutions that people would actually use.

LLMs still lack taste. They lack taste in design, in the visual arts, in literacy, in production code, etc. One of the hottest problems right now is to give LLMs taste. There are countless startups working on steering LLMs to be creative, giving LLMs more expert/domain-specific data, etc.

I doubt LLMs will ever truly have taste. The ability for us humans to find problems and to make things that people actually want is a result of a uniquely physical and emotional experience built over years of social interaction and living in our world. We notice problems because we are inconvenienced by them, or they make us upset.

WORK IN PROGRESS! check back in a few days to see updates…

standby Cups (When I’m Gone) Anna Kendrick