I think that one of the ways to ensure ‘the written word’ continues to have some value in the age of AI is to combine writing with speaking about what is written. In research, we do this routinely. We not only write papers but also talk about them. It is one of the better ways to know how well a person has understood something. It adds an extra dimension to the written communication.
When it comes to my physics blogs and articles, I have been doing it implicitly by combining my writing with some YouTube videos based on them. Writing still takes the central stage. It is just that I discuss what I have written so that it gets not only greater reach but also gives me room to explain things further.
Henceforth, I want to make this more explicit so that people appreciate what goes into writing. I will be discussing many of my writings on my physics YouTube channel – Humanizing Physics, which is linked below
Also, I see that many on social media tend to post AI-generated text without background research or ownership of the text. I do appreciate AI as a tool for research, but personally, I like writing and want to encourage people to value it both as an art and a craft. For me, writing still remains the most efficient way to think.
I feel that putting a face to the written word would lead to greater authenticity. Hence, this initiative.
The value of education is in understanding the world and responding to it. This understanding and responding are based on collective human knowledge. To obtain that knowledge is one of the main goals of education. Academia provides a platform, time and cohort to learn and interact with it. The result of education is the development of the mind and character. Of course, all this can be achieved outside academia, too. Just that it will have to displace, or at best, substitute for another activity. Whereas in academia, everything works ‘for’ education. Replacing academia may be an idea people are thinking of. The question remains what will be the form of its replacement. I don’t have a concrete answer to this, but I guess it will not be a YouTube channel. Nor will it be an AI chat box. These are great tools that can enhance education, but by themselves have primary goals that are different from what education tries to accomplish: a better human mind and a better human character. In its better form, academia’s explicit goals are exactly this. We humans, don’t have an alternative to academia (yet). Until then, we need to focus on the question: how can academia reinvent itself for the already bubbling AI era? Humans will be central to this transformation, of course, assisted by machines. The product of this transformation will be better human beings, and hopefully a better planet. We know the goal. The confusion is in the process and means to that goal.
1] Energy and resource expenditure are finite and constrained. AI companies dominate in LLM domains, and not (yet) in the physical 3D domain. That transition will perhaps be a greater leap, but probably the most challenging one. A great bridge built in America does not solve a problem in India, even when we have the complete knowledge and tools to build the bridge. Transferable solutions are easier to transfer in the software domain compared to hardware.
2] Human intervention is still needed. The definition of a scientific problem can be objective, but the pursuit of science (or maths) has to be anchored in the historical, philosophical and sociological setting. Humans will play a dominant role there, and a complete transition towards machines will be of lesser interest to humans.
We are in interesting and exhilarating times driven by AI info-tech, but we must pause to reflect as humans and ask what the implications are on our thinking and living. The answer will be interesting to study.
“The news today of progress on resolving the Navier–Stokes problem, one of mathematics’ great longstanding challenges concerning the equations that govern the flow of fluids, represents a milestone advance in human knowledge. This story began with Navier, Stokes, Leray, and Ladyzhenskaya and has culminated in the recent breakthroughs of Córdoba and Martínez-Zoroa, then — assisted by new technologies — Alpöge and Buckmaster, with the final steps taken by OpenAI mathematicians. The purpose of mathematics is human understanding, and this achievement, and the process that led to it, will bear fruit for a long time to come.” Ravi Vakil, President of the AMS John Meier, CEO of the AMS
The above snapshot is from OpenAI, which has claimed to have derived a new result in theoretical physics. What is it about, and how good are the claims? Below, I discuss them.
Let me start with some background. Except for the hydrogen atom, the nucleus of all elements in the periodic table consists of neutrons and protons. Neutrons and protons are made of quarks. Quarks interact through gluons. How do these gluons interact? This is a contemporary question.
In this particular case, the authors of the study say: “We’ve published a new preprint showing that a type of particle interaction many physicists expected would not occur can, in fact, arise under specific conditions. The work focuses on gluons, the particles that carry the strong nuclear force.”[1]
The interaction can be computed in terms of probabilities[2], and these probabilities depend on quantum mechanical amplitudes (also called scattering amplitudes). Finding these amplitudes requires a deeper knowledge of strong nuclear forces. Computing such amplitudes is expensive and requires a lot of effort. Physicists, under physical constraints, take a guess on which interaction is more probable and which is not. This study shows that one of the interactions that physicists thought was not probable turns out to be probable, but under specific conditions. “The preprint studies a central concept in particle physics called a scattering amplitude. A scattering amplitude is the quantity physicists use to compute the probability that particles interact in a particular way. …….One case, however, has generally been treated as absent (having zero amplitude)……..As a result, this configuration has largely been set aside. The preprint shows that this conclusion is too strong.”[1]
Of course, this has been possible using the brute force computational capability of the GPT 5.2 model, and it has come up with a particular formula that shows the amplitude to be probable and has further validated it with a formal proof. It is a methodological breakthrough, and the authors claim, “An internal scaffolded version of GPT‑5.2 then spent roughly 12 hours reasoning through the problem, coming up with the same formula and producing a formal proof of its validity.” [1]
I think it is a good development in computational physics and helps in calculating parameters that have relevance in finding probabilities of interaction in particle physics. Overall, my hunch is that it is an important step in computational physics.
Notes:
[1] OpenAI has put out an excellent summary of this problem (without jargon), and it needs basic physics, and the flow of text is good.
Below are a couple of paragraphs that caught my attention:
“Scientific understanding and scientific discovery are both important aims in science. The two are distinct in the sense that scientific discovery is possible without new scientific understanding….
…..to design new efficient molecules for organic laser diodes, a search space of 1.6 million was explored using ML and quantum chemistry insights. The top candidate was experimentally synthesized and investigated. Thereby, the authors of this study discovered new molecules with very high quantum efficiency. Whereas these discoveries could have important technological consequences, the results do not provide new scientific understanding.”
The authors provide two more examples of a similar kind, from different branches of science.
The authors conclude:
“Undoubtedly, advanced computational methods in general and in AI specifically will further revolutionize how scientists investigate the secrets of our world. We outline how these new methods can directly contribute to acquiring new scientific understanding. We suspect that significant future progress in the use of AI to acquire scientific understanding will require multidisciplinary collaborations between natural scientists, computer scientists and philosophers of science. Thus, we firmly believe that these research efforts can — within our lifetimes — transform AI into true agents of understanding that will directly contribute to one of the main goals of science, namely, scientific understanding.”
Nice.. thanks for sharing, will go through it. Although a lot of brute force seems to be passed off as understanding these days (brawn = brain?) I wonder if AI and ML of the varieties we have today are advancements in computing or intelligence?
My reply:
The computational capability is undoubtedly great, and probably the coding/software domain has been conquered, but there is a tendency to extrapolate the immediate impact of AI to every domain of human life, where even basic tech has not made an impact. That needs deeper knowledge of interfacing AI with other domains of engineering. Embedding AI in the virtual domain is one thing, but to put it in the real world with noise is a different game altogether. That needs interfacing with the physical world, and there is also an energy expense that doesn’t get factored into the discussion. It has great potential, and I’m eager to see its impact on the physical infrastructure. Parallelly, it is interesting to see how it has been sold in the public domain.
A gentle reminder: Digital infrastructure is not equal to physical infrastructure. The former is a smaller set of the four-dimensional space-time world we live in. AI-based tech is fantastic for an upgrade in digital infrastructure and has already made tremendous progress. But the real deal is in the physical domain. This also indicates where the future action is, and will be influenced by our understanding of physical sciences, including engineering domains beyond computer science. What we are witnessing in AI is probably the peak of Gartner’s hype cycle.
Recently, anthropic co-founder Jared Kaplan, who has a background in physics, made the following comment, which was circulated on X. Below is the excerpt:
Below is my response:
A Remarkable Human Being = Remarkable Attribute(s) + Human Being
The first term in the RHS can be replaced by AI, but not the second term, for the following reasons.
Machines, including AI, can surely change the way humans think, work and live, but it will be difficult to match human connection. A machine can enhance human life, but can it inspire a human life?
People inspire people. Ask a child or any adult who inspires them. It will generally be a fellow human being. Machines add value, but human beings represent a valuable life. We utilize the former, and get inspired by the latter. It is this inspiration that propels people forward to do things that may further turn out to be remarkable. This contribution is not easily quantified, but it is hard to gauge a human life without inspiration.
People like Ed Witten, Ashoke Sen and Terry Tao add value to humanity not only through their work and ideas, but their lives show that human beings can think and do something remarkable. It assures human beings that, individually, our species can do something good. Human beings derive meaning by interacting with fellow human beings and are inspired by the interaction. They also get inspired and draw meaning by studying people from the past. A human’s search for meaning and purpose is always in the background of other human beings. We are 8 billion plus, and it is hard to ignore each other.
It will be very unusual to find a serious student of theoretical physics who says I am inspired to live by ‘ChatGPT’.
Probably a young Kaplan, too, was inspired by a fellow human being! So, my question to Mr. Kaplan. Who inspired you to do physics?
Below is a small experiment with ChatGPT (GPT-4o, which is OpenAI’s most advanced model as of July 2025).
I asked for thoughts (from authentic sources) from Indian writers on the nature of the writing process (handwriting vs typewriting).
It confidently gave me quotations from Rabindranath Tagore, R.K. Narayan and U.R. Ananthmurthy. I was intrigued, and below is my conversation:
me: list the quotes from Indian writers above and give the exact sources
Chatgpt: I couldn’t find reputable sources for the exact quotes I previously suggested by Indian writers—so it’s safer to reconsider them.
This is something I have been observing with LLM chatboxes, especially when it comes to the sources that may not be mainstream. One may even say it is hallucinating, but after enquiring further, it said:
The quotes I initially provided appear to be popular paraphrases or second-hand interpretations, but none could be verified in primary texts or reliable interviews.
Then, it asked me if I wanted to use authentic sources to be presented to corroborate the initially mentioned sources. I said yes, and below was its reply:
Here are verified quotes from Rabindranath Tagore, with clear attributions to reliable sources—unfortunately, I couldn’t locate authenticated quotes for other Indian writers on writing methods
The quotes that it gave were peripheral to my initial question. It identified two to three quotations, but none of them were directly connected to my initial query.
There may be at least 2 hypotheses that I can draw from this, which need further testing:
ChatGPT and its kind may be getting trained more intensely on sources that are mainly anglophonic, and its geographical distribution of sources may be skewed. Online literary sources from countries such as India may not be as dense as, say USA, the UK or even some European countries. Will depositing more authentic sources online, including their translations, help the authentic discovery of information from countries such as India?
With the current developments and model training, there may already be a bias in the answers that LLM chatbots give. It may reinforce many viewpoints from Western repositories that may sometimes be disconnected or irrelevant to the user outside Western geographies. In that sense, new information is being built on old information. Are we entering a stage where data deposition asymmetry is creating an asymmetry of discovery?
I know these questions are not trivial to answer, but for LLM chatboxes to be authentic, they need to address questions with proper citations. I know some of them are trying to do that (eg, perplexity AI), but I find the links it provides for certain focused questions are not up to the mark.
My inference:
I am cautiously optimistic about the developments and achievements in source-based LLM interfaces, especially when you feed an authentic source (eg, NotebookLM).
But LLM chatboxes may be hyped when:
It comes to its capability of sourcing authentic information, and
The immediacy of replacements of existing knowledge systems.
LLM chatboxes should be treated as an experimental tool for utilitarian tasks where the information can be verified independently.
It is important to take the bottom line of ChatGPT seriously: ‘ChatGPT can make mistakes. Check important info.‘
