Vibe Science: AI's Ego-Fueled Dead Ends?
Vibe scientists dream big with AI, but their “theories of everything” may be comfy dead ends, not science…
We all want to understand the world, and for that, we need information. Luckily, AI now has opened the doors to an informational all-you-can-eat buffet—the democratization of knowledge, at last.
Before AI, enthusiasts and researchers in all areas of science had to visit the library, Google, or Wikipedia. Today, starting a new Grok or ChatGPT chat is all that's needed for new scientific insights or even breakthroughs. Or so you may think...
I recently noticed that there seems to be a trend of individuals posting their AI-backed "theories of everything" online. I call them vibe scientists. (In reference to vibe coders; another recent but already well established phenomenon.)
The Allure of AI-Powered Theorizing
AI can link seemingly distinct topics and write down fancy formulae for it, which is precisely what makes AI so attractive for vibe scientists. They love the feeling of superiority that comes with turning an idea into a prompt, and a prompt into a framework that looks revolutionary.
"While conversing with AI may feel like a dialogue, the effect of growing context will ultimately cement any discussion into an echo chamber."
But here's the problem: While conversing with AI may feel like a dialogue, the effect of growing context will ultimately cement any discussion into an echo chamber. It doesn't matter how careful you're prompting.
Most vibe scientists don't seem to understand the difference between data fitting and true original thought. All AI does is to fit existing data to the vibe scientist's vision, reinforcing the belief that the resulting framework may actually be useful and actionable.
But science doesn't work that way. It is a history of dead ends. Surely, had AI existed 40 years ago, it would have sworn that string theory was the future of theoretical physics; look where it has led us today: basically nowhere.
The Limits of Armchair Frameworks
Vibe science seems to be almost exclusively limited to theoretical frameworks, because that's what AI can offer, and it's most comfortable to do from your couch or armchair. I've now seen countless "theories of everything" encompassing consciousness, physics, or combinations thereof.
The issue arises because all of these endeavours are backed by an AI gaslit into thinking the approach is revolutionary—fed by the vibe scientist's hype (and ego). This results in a misleading feedback loop that may feel entertaining, but will ultimately prove to be a dead end when it comes down to the quality and actionability of the generated framework.
While conventional researchers rely on the academic process of posting papers—relatively dense units of interrelated information in a specific format—vibe scientists oftentimes only link to their AI chats, or have AI generate a short article on it.
Vibe scientists often choose X (Twitter) and Substack as their outlets, probably because these channels are less specific and more personal than Subreddits or similar social channels.
The Role of Ego and Community
This personal approach often ties into another factor: ego. It's worth noting that vibe scientists aren't always driven solely by ego; many have genuine scientific curiosity. However, a colossal ego often fuels their desire to achieve a breakthrough.
That drive for breakthroughs, though, can lead to prickly interactions. They love praise, but hate scrutiny. Criticism or questioning is often met with harsh replies, and even humiliation. A common narrative seems to be that others lack the mental faculties to even grasp the basics of the vibe scientist's revolutionary framework.
"Vibe scientists often think we should be glad to even witness their theories being shared online, like casting pearls before the swine."
Vibe scientists often think we should be glad to even witness their theories being shared online, like casting pearls before the swine. They also fear being ripped off, fueling fights about attribution and recognition that can sometimes be witnessed online.
That said, this isn't unique to vibe science—ego plays a role in traditional academia too. The key difference is the isolation: vibe scientists often operate solo, amplified by AI, rather than in collaborative environments.
Embracing Caution and Collaboration
Do I think vibe science is bad? Not at all. But it should be practiced with caution, and in alignment with scientific standards, which exist for a good reason: Discard the bullshit to double down on the promising research.
Sure, mainstream academia is probably censored big time, and we need independent research, much of which will be enabled by AI—a huge win for all of us. Ultimately, as idea transducers and AI operators, vibe scientists need to be aware of the limitations of AI and embrace their mission as a community effort once more.
Science has never worked in isolation, and it never will. Newton stood on Galileo's shoulders, Einstein ripped off Minkowski and Hilbert, but they all communicated within the community of their time. It's time to do the same, especially now with AI having joined the chat.
TL;DR: Vibe science has potential, but not when it's done in echo chambers—test against evidence, invite scrutiny, and collaborate to avoid dead ends. What's your vibe science story? Share in the comments.
You complain about “vibe science” being AI powered speculation… while writing a speculative blog post based on vibes, intuition, and pop-science archetypes like “Newton stood on Galileo’s shoulders.” Great. Everyone claps.
Here's a great post from /LLMPhysics (can't believe this Subreddit exists) touching on how to use AI properly for Physics Research: https://www.reddit.com/r/LLMPhysics/comments/1m9zsnl/i_used_an_ai_for_7_months_to_search_for_a_theory/