Mar 24, 2026 · 35m · latent-space
🔬There Is No AlphaFold for Materials — AI for Materials Discovery with Heather Kulik
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this podcast interview, MIT Associate Professor Heather Kulik discusses the real-world applications and fundamental limitations of machine learning in computational chemistry and materials science. She explains why materials discovery presents vastly harder quantum mechanical challenges than biological systems like AlphaFold, detailing her lab's breakthroughs in high-toughness polymers, carbon capture frameworks, and open-source molecular modeling tools.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. How this is scored →
The hosts as informed peer 3.5 Guest teaching 4.7 Guest disagreement 1.3 The hosts pushing back 0.7
speaking balance: gold is the hosts, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 13:40 Dismantling LLM chemistry hype
Kulik directly refutes the premise that LLMs replace chemists, highlighting that they only have Wikipedia-level competence and consistently fail basic chemical design constraints.
Hardest push from the hosts ▶ 13:16 Challenging the need to learn chemistryThe host poses a deliberately provocative counter-argument, pressing whether domain knowledge is obsolete when AI possesses apparent PhD-level grasp.
Biggest teaching moment ▶ 24:14 Explaining why AlphaFold cannot map to materialsKulik breaks down why protein folding with 20 amino acids is dramatically simpler than materials science, where diverse bonding across the periodic table makes generalization far harder.
The host holds their own ▶ 25:00 Host distinguishes ground state vs dynamicsThe host demonstrates domain grasp by articulating AlphaFold's focus on static ground states versus dynamical quantum mechanical systems in catalytic enzymes.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The hosts as informed peer | Guest teaching | Guest disagreement | The hosts pushing back | Why |
|---|---|---|---|---|---|---|
| Introducing Heather Kulik and Her Computational Chemistry Career | 2 | 3 | 0 | 0 | The host opens with an introduction and a broad question on Kulik's notable achievements. Kulik details her group's AI-driven polymer network discovery that experimentalists verified in the lab. | |
| Uncovering Quantum Mechanical Mechanisms in Polymer Dissipation | 4 | 4 | 1 | 1 | The host offers an analogy involving structural fuses in bridges to grasp the dissipation mechanism and brings up Kulik's early student projects. Kulik explains the quantum mechanical basis of molecular stabilization under stress. | |
| Multi-Objective Active Learning for Direct Air Carbon Capture | 4 | 4 | 0 | 0 | The host draws parallels between multi-objective materials discovery and drug development bottlenecks. Kulik explains seven-objective active learning campaigns for carbon capture and metal-organic frameworks. | |
| Integrating Machine Learning with Quantum Mechanical Modeling | 3 | 5 | 1 | 1 | The host asks how ML integrates with traditional quantum modeling. Kulik explains open-shell transition metal catalysis and feeding wave functions directly into neural networks to select quantum approximations. | |
| Domain Expertise vs. AI and the 22-Atom Ligand Benchmark | 3 | 7 | 4 | 2 | The host poses a provocative question on whether anyone needs to learn chemistry given ChatGPT. Kulik dismisses this idea by showing LLMs only possess Wikipedia-level knowledge and fail basic domain tasks like generating a 22-atom ligand. | |
| Identifying Dataset Bottlenecks and Neglected Chemical Regimes | 4 | 6 | 3 | 1 | The host inquires about community benchmarks analogous to CASP. Kulik critiques overhyped foundation potentials that fail on real systems and notes the lack of high-fidelity experimental ground truths. | |
| Overcoming Bottlenecks in Autonomous Labs and Device Processing | 4 | 4 | 1 | 1 | The host articulates the bits-to-atoms bottleneck and automation brittleness. Kulik emphasizes the neglected gap between materials design and device-scale manufacturing processes. | |
| Comparing Materials Complexity to Protein Folding and AlphaFold | 5 | 6 | 2 | 1 | The host probes the structural difference between materials and proteins. Kulik explains that materials possess diverse bonding across the periodic table, making the problem vastly more complex than AlphaFold's 20 amino acids. | |
| Extracting Literature Data and Navigating Scientific Ambiguities | 4 | 5 | 1 | 1 | The host raises concerns about ML models being biased toward historical literature. Kulik explains literature extraction methods and notes that published textual claims often contradict author graph data. | |
| Envisioning Public Cloud Labs and Machine-Learning-Ready Data | 3 | 4 | 0 | 0 | The host asks what infrastructure would best advance the field. Kulik outlines the promise of public automated cloud labs and machine-learning-ready reporting standards. | |
| Academic Chemistry in the Era of Tech Giant Compute | 3 | 4 | 1 | 0 | The host asks about the role of academic labs competing against massive corporate compute. Kulik argues that academics must focus on creative problem selection rather than brute-force computation, and shares open-source tools. |
Statements from this episode (16)
Kulik: AI Discovered Polymer Design That Made Plastic Four Times Tougher
“So we were able to screen with artificial intelligence a set of 1010 of thousands of materials where each individual experiment, if it were done in the lab, would have taken months to years. And through AI, we uncovered this sort of unexpected chemical phenome…”
Kulik: Polymer Toughness Relies on Previously Unobserved Quantum Mechanical Phenomenon
“But what we discovered was that there was a fully quantum mechanical phenomenon. There was really no way for us to predict this, you know, based on anything else, where the electrons just move around in a different way so that at this moment where the molecule…”
Kulik: ML's greatest chemistry potential lies in multi-dimensional challenges
“I think one of the areas where machine learning kind of just with what's out there right now has the most promising chemical sciences is in solving multi-dimensional challenges.”
Kulik: ML models deliver 100x to 1000x speedup per optimization dimension
“Usually, just even for a not so accurate machine learning model, you get, you know, at least a hundred to a thousand-fold speed up for every dimension you're optimizing over”
Kulik: Machine Learning Can Select Optimal Quantum Mechanical Approximations
“Not all quantum mechanical approximations are equal, and you can actually use ML models to kind of predict what the best approximation to use is, depending on the material studied.”
Kulik: LLMs consistently fail to generate a 22-atom ligand
“The thing I constantly do every time an LLM is updated is I just ask it, please design me a ligand that has, ah, 22 atoms. So the first time I've done that, there are many ligands out there that have 22 atoms, and then I say, I want it to bind to the metal wit…”
Kulik: Chemistry ML Datasets Only Cover Standard Organic Chemistry
“We have really good data sets out there for really boring chemistry. So we have, you know, probably even if you're not a chemist, you're familiar with organic molecule data sets, and Organic molecules binding to proteins. Those are the common data sets out the…”
Materials Project and Open Catalyst datasets rely on low-fidelity DFT calculations
“Materials project open catalyst project, these do provide good leaderboards, but some of the limitations that are the data comes from not very high fidelity density functional theory. So I'd say that's a second challenge is that we're all, all the smartest ML …”
Unnamed materials foundation model is only 5x faster than DFT and unreliable
“It's only in my hands the one I'm still not naming is only about five times faster than my fastest DFT calculation on a GPU, and it also doesn't work all the time.”
Kulik: Autonomous labs struggle with experiments that are easy for humans
“There are some types of experiments that, at least as of the last conference I went to on this, are really hard for autonomous
high throughput experimentation, but are really easy for a human and vice versa.”
Kulik: AI for materials is at 'ground zero' on manufacturing processing
“Most people who actually work on
Getting materials to the device scale, say something that would be in your television or something like that, is they will tell you that it's not just the material, it's the process.
And I think we're at ground zero.
We're nowh…”
Kulik: No Current ML Potential Robustly Models All Materials Bonding
“The challenge is that you have a lot more than 20 building blocks when it comes to materials and so there's lots of different ways to think about chemical bonding, and right now no potentials are really robustly encoding all of that bonding, especially with re…”
Kulik: Materials AI Models Can Fail Far More Catastrophically Than AlphaFold
“So it's just hard to know from experiment or from other computations if these types of models are correct, and they're certainly not correct across all of chemical space and I'd say they could fail more catastrophically than AlphaFold obviously fails, though t…”
Kulik: Literature breakdown temperatures from graphs frequently contradict authors' text descriptions
“One of the funniest things I think we noticed is that you can get the temperature at which a material will break down two ways. One, you can get it from the graph, and two, you can get it from what the authors say about how they interpret the graph. And those …”
Kulik: Machine-Learning-Ready Publishing Is Not Developed Across Materials Science
“Some research sub-fields are trying to do that, but it's not really developed across material science.”
Kulik: Academic Labs Must Avoid Problems Solvable by Brute-Force Compute
“For sure, Microsoft, Meta, those ones are kind of like the companies that have basically infinite resources, and as an academic, I don't have infinite resources, you know, but we have an interest in problems that, you know, haven't crossed the radar of those c…”