# LLMs Can’t Jump  ## Metadata - Author: [[Tom Zahavy, Google DeepMind]] - Full Title: LLMs Can’t Jump - Category: #articles - Summary: insert summary - My notes: - Summary: Einstein made scientific discoveries by using physical intuition and thought experiments, not just data or symbols. Modern AI can follow rules but cannot create new fundamental ideas without sensory experience. This shows AI still cannot fully replicate human creativity in science. - URL: https://readwise.io/reader/document_raw_content/415569447 ## LLM Chats ## NotebookLM ## LLM Audio ## Highlights > However, this capability comes with a critical caveat. An AI can deduce the consequences of "The Equivalence Principle" only if those concepts are provided as inputs. As Einstein noted, logical thinking is limited to connections between concepts; it cannot generate the concepts themselves from raw sensory data. The 1913 derivation failed not because the logic was flawed, but because the axioms were incorrect. This leads us to the fundamental bottleneck: what cognitive process allowed Einstein to generate the "Equivalence Principle" in the first place? To understand this, we must look beyond logic to the mechanism of the "Jump" (). ([View Highlight](https://read.readwise.io/read/01kh0s1b90s9sat59k3c63a7pc)) > ARC captures the *logical* leap, it misses the *manipulative* component—the physical sensation and embodied simulation that drove Einstein's insight. ([View Highlight](https://read.readwise.io/read/01kh0s67g1tw078y560hscxhk2)) > "The words or the language, as they are written or spoken, do not seem to play any role in my mechanism of thought." ([View Highlight](https://read.readwise.io/read/01kh0s904hx1eyj36p1medzpr7)) > these symbols in physical reality. They operate as highdimensional "Chinese Rooms" [(Harnad,](https://readwise.io/reader/document_raw_content/415569447/#page-7-4) [1990)](https://readwise.io/reader/document_raw_content/415569447/#page-7-4) , manipulating the language of physics without access to the physical referents that give that language meaning. This limitation prevents the AI from making the Abductive Jump ( → ). While Einstein could ground his axioms in the physical experience of a falling body, an LLM is confined to the logical deduction of existing texts. ([View Highlight](https://read.readwise.io/read/01kh0sat9vntd7x84v0n7zw26j)) > Finally, we emphasize that this proposal is specifically tailored to the physical sciences, where the object of study is external material reality. In abstract domains such as Mathematics or Computer Science, the Sense Experience () may be grounded in highdimensional topology or have other goals such as generality or minimality. While the necessity of the Abductive Jump remains universal, the nature of the simulation must be adapted to the ontology of the discipline: for physics, the substrate is the world; for mathematics, it is the abstract landscape of formal systems. ([View Highlight](https://read.readwise.io/read/01kh0sn9k0ma6b47m8tp0b75c3)) > Lastly, its important to note that Einstein relied on his Physical Prior, using the sensation of gravity to prune the search space of possible axioms. However, manipulative abduction extends beyond physics. Historical scientific revolutions are often driven by strong, pre-symbolic intuitions—whether Kepler's Neoplatonic belief in the centrality of the Sun or the 'objective anger' that drove Marx's modeling of capital. To automate invention, we may need systems that do not just simulate the world, but hold strong beliefs or priors about how that world should be structured, using simulation to test those specific intuitions. ([View Highlight](https://read.readwise.io/read/01kh3ra2np6ctx44bh9fsx6zt3)) > This limitation is visible even in the apex of today's automated discovery systems. Agents such as Sakana's AI Scientist [(Lu et al.,](https://readwise.io/reader/document_raw_content/415569447/#page-8-11) [2024)](https://readwise.io/reader/document_raw_content/415569447/#page-8-11) and Google DeepMind's AlphaEvolve [(Novikov et al.,](https://readwise.io/reader/document_raw_content/415569447/#page-8-12) [2025)](https://readwise.io/reader/document_raw_content/415569447/#page-8-12) demonstrate the immense power of mechanizing scientific loops and evolutionary optimization. However, they highlight the very abductive gap we identify. The AI Scientist recombines existing symbolic concepts to optimize metrics—a sophisticated "Chinese Room" operation that lacks the sensory grounding to invent axioms without symbolic precedent. Similarly, while AlphaEvolve excels at optimization within a fixed framework, it relies on a gradient; Einstein, by contrast, had no error signal from Newtonian mechanics to drive his discovery. These systems lack the embodied world model required to perform the counterfactual physical simulations that drive the abductive Jump to entirely new paradigms. ([View Highlight](https://read.readwise.io/read/01kh3vrpyce9eezpe9s32x3p6z)) > Unlike deduction, which guarantees truth, or induction, which finds pattern that generalize in data, abduction is a creative leap that invents a cause for a singular phenomenon. Crucially, Einstein achieved this via embodied simulation—using thought experiments to ground abstract symbols in physical sensation—enabling him to formulate axioms where no symbolic data previously existed. We argue that while Large Language Models have mastered the inductive compression of data and the deductive verification of theorems, they are structurally incapable of the abductive 'jump' required for scientific invention. We posit that this creative leap demands not just better language processing, but the integration of physically consistent World Models that ground abstract symbols in sensory simulation. ([View Highlight](https://read.readwise.io/read/01kh0r077exjdnxytd66dh44nc)) > Einstein did not discover General Relativity by searching over symbols; he discovered it by simulating the sensual experience of a falling observer. ([View Highlight](https://read.readwise.io/read/01kh0sjn39bpbbxcpswvmty1ra))