Google has published a paper that might end the transformer era. For the last 7 years, every major AI, ChatGPT, Claude, Gemini, has been built on the exact same architecture: The Transformer. But Transformers have a fatal flaw. To remember context, they have to process every single word against every other word. It’s called quadratic complexity. As your prompt gets longer, the compute cost explodes. The alternative is the old-school RNN (Recurrent Neural Network). RNNs are incredibly cheap and fast, but they have a fixed memory size. If you give them a long document, they get amnesia. Until today. Google researchers published Memory Caching: RNNs with Growing Memory. And it fixes the biggest bottleneck in AI. Instead of an RNN having a fixed, rigid memory that constantly overwrites itself, Google gave it a "save" button. The technique allows the RNN to cache checkpoints of its hidden states as it reads. The memory capacity of the RNN can now dynamically grow as the sequence gets longer. They built four different variants, including sparse selective mechanisms where the AI actively chooses exactly which checkpoints matter most. The results rewrite the rules of efficiency. On long-context understanding and recall-intensive tasks, these new Memory-Cached RNNs closed the gap with Transformers. They achieved competitive accuracy without the explosive, quadratic compute cost. It perfectly bridges the gap between the cheap efficiency of an RNN and the massive capability of a Transformer. We have spent billions scaling Transformers because we thought they were the only way an AI could remember a long conversation. But Google just proved we don't need to process the whole history every single time. We just needed a smarter cache.

🚨BREAKING: Researchers built an AI that designs better AI than humans can. It discovered 105 new architectures that beat human-designed models. Nobody guided it. It taught itself. The paper is called "ASI-Evolve: AI Accelerates AI." Published this week by researchers at Shanghai Jiao Tong University. Fully open-sourced. And what it demonstrates should stop every AI researcher cold. They built a system that runs the entire AI research loop on its own. It reads scientific papers. It forms hypotheses. It designs experiments. It runs them. It analyzes the results. Then it uses what it learned to design better experiments. Over and over. Without human intervention. They pointed it at neural architecture design first. Over 1,773 rounds of autonomous exploration, the system generated 1,350 candidate architectures. 105 of them beat the best human-designed model. The top architecture surpassed DeltaNet by 0.97 points. That is nearly 3 times the gain of the most recent human-designed state-of-the-art improvement. Humans spent years to get 0.34 points. The AI got 0.97 on its own. Then they pointed it at training data. The AI designed its own data curation strategies and improved average benchmark performance by 3.96 points. On MMLU, the most widely used knowledge benchmark, the improvement exceeded 18 points. Then they pointed it at learning algorithms. The AI invented novel reinforcement learning algorithms that outperformed the leading human-designed method GRPO by up to 12.5 points on competition math. Three pillars of AI development. Data. Architecture. Algorithms. The AI improved all three by itself. Then they tested whether what the AI built actually works in the real world. They applied an AI-discovered architecture to drug-target interaction prediction. It achieved a 6.94 point improvement in scenarios involving completely unseen drugs. The AI designed something that works better than human experts in biomedicine. This is the first system to demonstrate AI-driven discovery across all three foundational components of AI development in a single framework. The recursive loop is now closed. AI is building AI. And it is already better at it than we are.

It is the deepest honor to have been joined by Michael Levin (@drmichaellevin), Victoria Klimaj, Zahra Sheikhbahaee (@zah_bah), Dalton Sakthivadivel (@DaltonSakthi), Adeel Razi (@adeelrazi), David Ha (@hardmaru), Nick Hay, Kevin Schmidt, Irina Rish (@irinarish), David Krakauer (@sfiscience), Melanie Mitchell (@MelMitchell1), Samuel Gershman (@gershbrain), and Joshua Tenenbaum in organizing this special issue of the Royal Society’s (@RSocPublishing) Philosophical Transactions A: “World models, A(G)I, and the Hard problems of life-mind continuity: Toward a unified understanding of natural and artificial intelligence” royalsocietypublishing.org/r… This collection was motivated by a question with far reaching implications, ranging from the fundamental nature(s) of mind to choices that may determine the future of our civilization/species: what kinds of world modeling capabilities are likely to be realized by which kinds of minds and what world might we be in with respect to increasingly advanced artificial intelligences? Will the scaling and refinement of present approaches result in AI with human-like (and beyond) cognitive abilities, or do we need radically different paradigms that more closely follow the principles of natural intelligence? Learning “world models” to predict/compress information may be how biological learners so efficiently learn (to learn) to achieve goals and generalize that knowledge across a broad range of task environments. World models may also be useful for reverse-engineering forms of “System 2” cognition, or the self-reflexive, deliberate, multi-step reasoning associated with cognitive capabilities that may be unique to humans. Predictive models that reflect how the world may be causally modified by actions allow agents to adaptively control their behavior with flexibility and context-sensitivity. Spatiotemporally and causally coherent models of the physical world may not only be the key for creating AIs that we can rely on for real-world deployment, but may even be the (dynamic) core of conscious cognition. The contributions to this special issue consider the varieties of world models worth modeling from diverse points of view: Douglas Hofstadter explores whether sufficiently coherent self-referential world modeling could ground meaning, consciousness, and a genuine “I” in future AI systems. David Krakauer (@sfiscience), Melanie Mitchell (@MelMitchell1), and John Krakauer (@blamlab) examine the principles of emergent intelligence from a complex systems perspective. Alexander Ku (@alex_y_ku), Declan Campbell, Xuechunzi Bai (@baixuechunzi), Jiayi Geng (@JiayiiGeng), Ryan Liu (@theryanliu), Raja Marjieh (@RajaMarjieh), R. Thomas McCoy (@RTomMcCoy), Andrew Nam, Ilia Sucholutsky (@sucholutsky), Liyi Zhang (@LiyiZhang_Leo), Jian-Qiao Zhu (@JQ_Zhu), and Thomas Griffiths (@cocosci_lab) argue for using the tools of cognitive science to understand and evaluate LLMs across multiple levels of analysis. Evelina Leivada (@EvelinaLeivada), Gary Marcus (@GaryMarcus), Fritz Günther, and Elliot Murphy (@ElliotMurphy91) test whether LLMs deeply understand language and the “world behind words,” or primarily learn surface statistical regularities. Pedro Tsividis (@ptsividis), João Loula, Jake Burga, Juan Pablo Rodriguez, Sergio Arnaud, Nate Foss (@_npfoss), Andres Campero, Ajay Subramanian (@ajaysub110), Thomas Pouncy, Samuel Gershman (@gershbrain), and Joshua Tenenbaum introduce a theory-based meta-learning architecture inspired by the remarkable flexibility and efficiency of human cognition. Eunice Yiu (@eunice_yiu_), Kelsey Allen, Shiry Ginosar (@shiryginosar), and Alison Gopnik (@AlisonGopnik) explore empowerment, controllability, and causal reasoning as means of understanding the remarkable learning abilities of both child and adult minds. Nadav Amir, Stas Tiomkin, and Angela Langdon investigate how goals shape the structure of experience and how the world modeling abilities of natural intelligences may be inseparable from values. Vickram Premakumar, Michael Vaiana, Florin Pop (@FlorinPop17), Judd Rosenblatt (@juddrosenblatt), Diogo Schwerz de Lucena, Kirsten Ziman, and Michael Graziano show unexpected benefits of self-modeling as an inductive bias and regularizer for training artificial agents. Hanlin Zhu, Baihe Huang, and Stuart Russell analyze why model-based reinforcement learning may fundamentally outperform model-free approaches in representational efficiency. Bradly Alicea (@balicea1), Morgan Hough (@mhough), Amanda Nelson, and Jesse Parent (@JesParent) revisit fundamental cybernetic principles of regulation, adaptation, and world modeling across a wide assortment of complex adaptive systems. Francesco Sacco (@FrancescoSacco1), Dalton Sakthivadivel (@DaltonSakthi), and Michael Levin explore topological constraints on self-organization and suggest that biological systems maintain long-range coherence in ways that are fundamentally different from current transformer architectures. Georg Northoff (@NorthoffL), Yasir Catal, and Samira Abbasi examine how biological intelligence may depend on capabilities for flexible “inner time” to ensure adaptive alignment between the dynamics of system and world. Nicolas Rouleau (@DrNRouleau) and Michael Levin explore whether theories of consciousness generalize beyond brains to unconventional embodiments and living systems more broadly. Benjamin Lyons and Michael Levin investigate economies and collective intelligence as systems coordinated by “cognitive glues” in the form of shared models of scarcity and value. Katherine Collins (@katie_m_collins), Umang Bhatt (@umangsbhatt), and Ilia Sucholutsky (@sucholutsky) consider “Rogers’ paradox” to demonstrate ways in which collective learning is impacted by different kinds of human-AI interactions. Ruairidh Battleday (@RMBattleday) and Samuel Gershman (@gershbrain) distinguish between the “easy” and “hard” problems of science, and describe how while current AI systems demonstrate powerful narrow forms of optimization with respect to well-defined inference-spaces, further developments are needed for achieving capabilities for novel scientific discovery. Fritz Breithaupt (@FritzBreithaupt) explores narrative world models and the roles of uncertainty and transformative experiences in natural intelligences, suggesting that coherent agency may depend on better understanding human-like meaning-making. Taken together, these diverse perspectives suggest that while LLMs can clearly learn powerful generative models of language, they likely do so without having world models of sufficient spatiotemporal and causal coherence to achieve human-like reasoning abilities, capacities for generating subjective conscious experiences, or pathways to realizing artificial general superintelligence. However, by further developing world modeling architectures, we may eventually be able to create forms of intelligence that recapitulate the remarkable flexibility and generality of human intelligence. Finally, enhanced (e.g. more coherent/integrated) world models may not only afford expanded capabilities, but could potentially help ensure that increasingly powerful AI systems achieve both inner and outer alignment with human(e) values.

"You’re not talking to someone who woke up a loser” - Jensen Huang Jensen nearly lost his composure during a heated debate about selling chips to China, despite showing tremendous patience in response to the pushback.

Same, I have a similar setup. A mix of Obsidian, Cursor (for md), and vibe-coded web terminals as front-end. Since I do a podcast, the number/diversity of research interests is very large. But the knowledge-base approach has been working great. For answers, I often have it generate dynamic html (with js) that allows me to sort/filter data and to tinker with visualizations interactively. Another useful thing is I have the system generate a temporary focused mini-knowledge-base for a particular topic that I then load into an LLM for voice-mode interaction on a long 7-10 mile run. So it becomes an interactive podcast while I run, where I ask it questions and listen to the answers to learn more. Anyway, heading out for a run now, thanks for the write-up 👊