Sabine Hossenfelder on AI, bad physics, and why science needs reform
- Colin Hunter
- Jul 22
- 5 min read
Updated: Aug 12
She’s one of the internet’s sharpest scientific voices—equal parts physicist, critic, and communicator. In this candid conversation, Sabine Hossenfelder reflects on AI, the flood of low-impact theory papers, and how a scientific culture ripe for reform could finally be ready for change.
Sabine Hossenfelder doesn't pull punches. Nor does she flinch when punched.
She has become one of the internet’s favourite scientific commentators – a willfully independent theoretical physicist unafraid to criticize the shortcomings and decadences of her field.
She skewers the publish-or-perish treadmill of academia, the pursuit of esoteric questions disconnected from reality, and lately, the proliferation of algorithmically generated “AI slop” in science.
“Yeah, I think it's definitely going to be a huge change," she tells FirstPrinciples in an exclusive interview from her Munich studio. “We're seeing the beginnings of it, and it will definitely have some good parts and bad parts.”
That dichotomy of AI’s promise and peril is core to Hossenfelder's evolving view, which she regularly shares with her 1.7 million YouTube followers. Viewers of her channel get a relatively rare look at a renowned theoretical physicist learning, thinking, and sharing ideas in real time. Her videos drop five times a week, which means she puts her neck out far more frequently than she possibly could through the peer-review process or working for a university.
Hossenfelder earned her PhD in physics from Goethe University in Frankfurt, worked at top institutions like the Perimeter Institute and Nordita, and has published more than 80 peer-reviewed papers on quantum gravity, cosmology, and high-energy theory. Her popular books Lost in Math and Existential Physics exhibit the same bluntly sharp style as her YouTube series, where she blends technical fluency with bracing straightforwardness.
AI and the future of physics publishing
The danger of AI in the short term, she says, is that academic journals already groan under the weight of low-impact theory papers. AI, she says, will spawn a flood of recycled ideas in which original ideas will drown. “We already have too much of this stuff anyway. So now we're about to get a lot more."
She envisions AI exacerbating a process she already sees in physics: generate some meaningless numbers with a simulation, invent a new particle with no reason to exist, slap it into a standard format, and then hit publish. “It's already pretty much an automated process. You fiddle around with the parameters until it kind of works, submit it, and it'll almost certainly be accepted because that's the current standard practice."
What concerns her most is that these papers don’t move science forward. “If you want more nonsense, yeah, then you can use GPT. If you want to actually make progress, it's not so easy."
Sabine Hossenfelder on AI’s promise (and limits)
But Hossenfelder isn’t an AI doomer. She uses AI tools herself and often marvels at their efficiency. “They're good at reading a lot of stuff very quickly and figuring out how it fits together." She recently tested that by feeding an LLM a confusing claim from a decades-old paper. “It says, ‘Oh, it's broken. Probably because XYZ.' And I look it up, and it's actually correct."
That's the upside, as she sees it: AI as a sort of supercharged librarian, able to connect dots across a literature base that no human could fully digest. “There’s just too much literature. It’s just not possible for physicists to keep track of all of this."
Still, she’s skeptical of AI’s ability to actually innovate in a meaningful sense. “They're not truly creative.”
Navigating the Paper Flood: How AI Could Reshape Scientific Publishing
Hossenfelder believes the root problem is deeper than bad algorithms. It's the state of theoretical physics itself. “A lot of research in foundations of physics is more or less nonsense," she said. “We’re actually not doing it the right way."
As a concrete example, she described her work exploring dark matter as a type of superfluid. The mathematical terrain was unfamiliar, and few people were working at that interdisciplinary junction. “There are a lot of connections to be made between these two areas," she said, “but there are so few people working at this intersection that it isn't happening."
Could AI help make those connections? Maybe, she said, but first we’ll have to make it through the “valley of despair."
“At the moment, I think things are going to get worse," she said. “But sometimes things have to get really bad before they can get better."
She hopes the coming deluge of low-effort, AI-generated research will finally force change. “It'll make the problem that we have in the foundations of physics—this pointless stuff that gets published—much more apparent. And it'll force people to finally do something about it."
Groupthink and the ‘bubble’ of modern physics
She’s equally candid about groupthink in academia and how breaking those bubbles could unlock more diverse, creative lines of inquiry: “We all talk too much to each other, so we all kind of think alike, we all think about the same things, we learn the same techniques."
She applies the same scalpel to string theory. Once hailed as a potential “theory of everything," it has, in her view, bloated into a sprawling “bubble” of self-justifying complexity. “It just got worse and worse. And every time something like that happened, they fixed the problem. So at some point, you're like, I don't think it's this elegant, simple thing you told us it'd be."
Provocation with purpose: YouTube as a platform for physics reform
That kind of piercing analogy is her trademark of her YouTube channel, which includes such titles as:
Hossenfelder knows she’s being provocative with titles like that. She’s throwing jabs at raw nerves. She’s doing so in hopes that viewers—physicists included—will confront difficult truths and work harder.
Her reach (and her willingness to engage deeply and frequently) has made her a bridge between the scientific community and people hungry for honest, intelligent commentary.
“I think I am psychologically equipped for controversy. I don’t know why. It just doesn’t bother me. I don’t understand why people get so upset. I believe I’m just pointing out things that they should be thinking about.”
Colin Hunter is a science communicator, filmmaker, and contributor to FirstPrinciples. He previously led the communications teams at the Perimeter Institute for Theoretical Physics and the Institute for Quantum Computing (IQC) at the University of Waterloo.
