No Quantum Gravity, No Real AGI
A physicist sets a high bar for AGI: it must solve quantum gravity. This conversation also clarifies the real relationship between quantum materials and AI—once the synthesizability filter is applied, the number of AI-discovered materials may drop to zero.
The argument · tap a timestamp to hear it
Too much learning paralyzes; naivety lets you ask original questions
The hardest step from student to researcher is not absorbing knowledge but asking original questions that push the boundary of knowledge. Gopakumar describes this leap as 'naivety plus experience': learning too much can paralyze you; you need to retain a bit of naivety to dare ask questions that seem impossible to tackle. A mentor's role is to teach you to break big questions into small steps you can advance today without losing the ultimate goal. His mentor David Gross excelled at this—according to Gopakumar, Gross's first student produced work that won a Nobel Prize for both mentor and student, and Witten also won a Fields Medal. Now Gopakumar replicates the same method with his own students: he assigns the first one or two problems, even ones he himself has no answer to, and accompanies students through them until their third or fourth year, then lets them ask their own questions.
— Rajesh GopakumarSpacetime is not a background but an emergent approximation
Using 'water is made of discrete atoms' to understand string theory: the continuous spacetime we touch is likely a low-resolution approximation of some deeper structure at macroscopic scales. Gopakumar stresses this is not pure speculation—quantum mechanics itself does not allow treating spacetime as an infinitely divisible smooth background; plus black holes (which Hawking and others found to have many anomalous properties) and the failure at the Big Bang singularity force physicists to answer 'where does spacetime come from?' String theory happens to provide a natural framework: spacetime emerges from something more fundamental, and gravity and the known interactions can fit inside. This explains why he regards 'solving quantum gravity' as a very high bar for intelligence: it requires not just solving problems but building new conceptual frameworks.
— Rajesh GopakumarMaterials should be designed in reverse: specify properties first, then find structure
First, a long-term yardstick: quantum mechanics was initially only for understanding atoms, but now the devices it underpins account, by his estimate, for about 35%-40% of US GDP, and the commercial ripple only became obvious four to five decades after the theory appeared. He calls the current phase the second quantum revolution: quantum materials like graphene, discovered 'by accident,' are already being commercialized. The next step is using AI for inverse materials design—first define what properties a material should have, then search for what structure can achieve them and how to synthesize it, bypassing traditional trial-and-error chemistry. He also reveals a conversation with Demis Hassabis: DeepMind's next holy grail is finding a room-temperature superconductor, and Hassabis told him at the outset not to talk about quantum gravity.
— Rajesh GopakumarGetting stuck is the norm; keep three or four problems in play
Being stuck is not the exception but the daily life of a researcher; especially during the transition from student to researcher, being stuck can feel crushing. Gopakumar's first strategy is portfolio hedging: work on three or four problems simultaneously; if one dies, switch to another, and by probability at least one will move half a step. The second trick is to teach—when students' eyes light up, that satisfaction can pull you out of a low point; he has often found that 'after a lecture, a stuck problem suddenly opens up.' The third trick is to deliberately learn something new you have always wanted to learn but never had time for, to gain a new frame of reference before returning to the old problem.
— Rajesh GopakumarTalent is multiplicative, not additive; the shortage is never headcount
Why does top science happen in a few places rather than uniformly distributed? Gopakumar offers a mechanism: being in the same place, conversing, using each other as sounding boards—even without formal collaboration—keeps the brain in a stimulated state; when collaboration does happen, the different skills two people bring multiply rather than add. He also uses a physics metaphor: photons in an ordinary light bulb are chaotic, while photons in a laser are coherent with each other; once a group forms coherence, it can produce a lasing effect like a laser. He says many Indian institutions' performance does not match their talent; what is missing is not headcount but this resonant atmosphere. Therefore ICTS's founding principle is simple: hire only first-rate people, give them resources and freedom, and do not let them lower their ambitions.
— Rajesh GopakumarAI can accelerate research but cannot make conceptual leaps
AI just won two Nobel Prizes in a row, but Gopakumar splits 'AI for science' and 'science for AI' into two things: the former he acknowledges as an overdrive gear—AI is accelerating a narrow class of tasks in research; but the 'theory' AI does today is far from his conceptual standard, as the best physics relies on conceptual leaps rather than pattern recognition. Conversely, he is more optimistic about science for AI—among those who discovered scaling laws at Anthropic, some have physics backgrounds; Hopfield himself is a statistical physicist who used simple physics models to simulate the brain. Along this path, physicists may join computer scientists and neuroscientists to move AI from language models to physics-inspired world models. This distinction is worth remembering.
— Rajesh GopakumarAfter the synthesizability filter, AI's new materials number zero
The most sobering cold water in the whole talk came from a Q&A: Google claims AI has discovered a large number of new materials, but after filtering out feasibility and synthesizability, Gopakumar says 'I think left with zero'—zero. His attitude is not to dismiss AI materials science (someone in the audience cited Microsoft's example of using a DFT model with only 385,000 parameters, bypassing traditional hybrid methods), but to remind that the whole chain is still long: after predicting crystal structures, you still need to grow them and scale them up, and each step can filter out surprises. He says India should also bet on this chain. This statement is a useful counter-calibration to the 'AI discovers new materials' narrative.
— Rajesh GopakumarIn their own words · checked verbatim
I think I'll believe that there is AGI when it solves quantum gravity.
Rajesh Gopakumar0:02
he was asking me what do you think are the big unsolved physics question and he said don't tell me about quantum gravity
Rajesh Gopakumar40:45
a laser beam is much more powerful than your regular light bulb. The laser beam has a coherent set of photons and that's the Bose Einstein effect.
Rajesh Gopakumar55:08
the point was that you you you are not just a cog in the wheel like you are might be in some of the welloiled machines of Stanford or Princeton or Harvard or some places.
Rajesh Gopakumar58:10
as a nation I mean we are a many millennia old civilization we should be thinking in in in terms of the long term investments and that's what I think is going to to to lift the country.
Rajesh Gopakumar1:10:19
I think of it as a little genie on my uh table uh which uh I mean I never used to code.
Rajesh Gopakumar1:26:38
Figures
| Share of US GDP from devices using quantum mechanics | about 35%-40% | 35:44 |
| Number of naturally occurring elements | 92 | 8:17 |
| Non-US nationals among Princeton's same-year theoretical physics PhD students | 10 out of 18 | 13:24 |
| Number of problems Gopakumar suggests researchers work on in parallel | 3-4 | 50:59 |
Glossary
- Quantum gravity
- The theoretical goal of unifying general relativity and quantum mechanics; no complete theory exists yet, and string theory is a candidate framework.
- String theory
- A framework assuming fundamental particles are different vibration modes of one-dimensional strings, attempting to unify gravity and the other forces and explain the emergence of spacetime.
- Bose-Einstein condensation
- A state where many bosons occupy the same quantum state at low temperatures; used here as a metaphor for the lasing effect when talented people resonate.
- Synthesizability
- Whether a computationally predicted new material can actually be synthesized; the most likely failure point in AI-driven materials discovery.
How to listen
Founders and investors working on AI for science, and engineers and scientists torn between joining a big platform or doing hard-core research with a startup mindset.
The first 25 minutes of academic biography can be fast-forwarded; start from the middle if you want the arguments.