The INTO THE IMPOSSIBLE Podcast - #239: The Computer EXPERT That Just Solved Google's Hardest Challenge

🔖 Titles

1 / 1

1. Can AI Discover New Laws of Physics? Rose Yu on Scientific AI, Traffic and Pandemic Forecasting 2. Rose Yu on AI as Scientist: From Traffic Jams to Fundamental Physics and Pandemic Predictions 3. The Next Einstein? Rose Yu Explores AI-Created Scientific Theories and Their Real World Impact 4. How AI Is Accelerating Scientific Discovery with Rose Yu: Symmetry, Traffic, and Supercomputers 5. Machines and Minds: Rose Yu on AI’s Role in Physics, Epidemics, and the Future of Science 6. Building AI Scientists: Rose Yu Reveals How Data-Driven Models Uncover New Physical Laws 7. Rose Yu: AI That Discovers Symmetries, Improves Traffic, and Forecasts Pandemics 8. Can AI Truly Create? Rose Yu on Machine-Generated Hypotheses, Physics, and the Human Element 9. From Game GPUs to Scientific Breakthroughs: Rose Yu on AI’s Transformative Power 10. Humanity, Science, and AI: Rose Yu on Trust, Discovery, and the Future of Research

💬 Keywords

1 / 1

AI scientific discovery, artificial intelligence and emotions, deep learning models, Large Hadron Collider data, symmetry detection, Lorentz invariance, Einstein Equivalence Principle, generative AI, mathematical theorem generation, hypothesis generation, molecule discovery with AI, GPU architecture, matrix multiplication, physics simulations, space-time discretization, AI limitations, data-driven simulators, climate modeling, traffic forecasting, Google Maps traffic prediction, diffusion convolutional neural networks, pandemic forecasting, epidemiological modeling, fusion energy research, quantum computing in AI, trustworthy AI, AI safety and guardrails, multimodal foundation models, human-AI collaboration, future of education with AI, interdisciplinary research

💡 Speaker bios

1 / 3

Brian Keating is a renowned physicist and professor at UC San Diego whose work has garnered both acclaim and admiration among colleagues and the broader scientific community. Known for tackling profound and provocative questions on the nature of the universe, Keating draws inspiration from great thinkers like Albert Einstein. He often explores the limits of human and artificial intelligence in physics, pondering whether an AI physicist could ever achieve insights as revolutionary as Einstein’s “happiest thought”—the equivalence principle that redefined our understanding of gravity. Through his research, teaching, and thought-provoking discussions, Keating continues to advance the field, earning the recognition he so richly deserves.

ℹ️ Introduction

1 / 1

Welcome back to the INTO THE IMPOSSIBLE Podcast! In this captivating episode, host Brian Keating sits down with Professor Rose Yu, a computational physicist at UC San Diego with a brilliant track record in artificial intelligence and cross-disciplinary research. From working on AI-driven traffic forecasting models—now powering Google Maps—to leading pandemic predictions that ranked number one nationally, Dr. Yu is at the cutting edge of how machine learning intersects with the real world. But the conversation doesn’t stop at practical applications. Together, Brian and Rose dive deep into the philosophical and technological frontiers of AI. Can artificial intelligence ever replicate the creative leaps of minds like Einstein? Will machines someday discover new laws of physics on their own—or even experience something akin to “shivers down the spine”? Rose reveals how her team’s AI models have uncovered fundamental symmetries in particle physics, once thought discoverable only through human genius, and explores both the promise and limitations of machine discovery, from deep learning hardware to the future of quantum computing. The episode also takes you behind the scenes of Rose’s cross-disciplinary ventures—connecting the dots between traffic jams, climate science, nuclear fusion, and pandemic response. Sprinkled throughout are reflections on the role of scientists in the age of AI, the future of education and academia, and the ethical guardrails we must build as intelligent machines move ever closer to collaborating in scientific discovery. Prepare to rethink what’s possible when human creativity teams up with artificial intelligence, and gain fresh insights into a future where the boundaries between natural and artificial intellect are rapidly blurring. Don’t miss this fascinating exploration into the next frontier of knowledge!

📚 Timestamped overview

1 / 2

00:00 "AI Unveils Hidden Universe Patterns"

04:41 "GPUs Drive Efficient AI Computing"

08:21 Disentangling AI Models from Hardware

12:51 Deep Learning Transforms Traffic Forecasting

16:39 Fluid Simulation: Eulerian and Lagrangian

20:24 Analyzing Traffic Data Challenges

22:51 Timing Optimization in Apps

24:53 AI's Role in Pandemic Response

28:30 Driven to Impact Traffic and Pandemic

33:33 Superintelligence: Friend or Foe?

34:13 AI Safety and Guardrails

40:13 AI as a Scientific Assistant

42:19 Open Source Molecular Generation App

47:23 Evolving Role of Educators

50:23 "Embrace Risk, Follow Your Passion"

52:42 AI Future: Collaboration or Replacement?

❇️ Key topics and bullets

1 / 1

Absolutely! Here’s a comprehensive breakdown of the main topics covered in this episode of The INTO THE IMPOSSIBLE Podcast with Professor Rose Yu, along with sub-topic points under each primary topic: --- ### 1. Introduction to Rose Yu and the Capabilities of AI in Scientific Discovery - Rose Yu’s background and achievements - The concept of AI discovering physical symmetries from data (e.g., Lorentz invariance) - The provocative question: Can AI replicate the creative leaps of Einstein and other geniuses? ### 2. Emotional and Creative Capacity of AI - Can AI experience emotions or “happy thoughts” like humans? - Distinguishing between AI’s lack of emotion and its ability to create new content - Examples: AI creating new theorems, molecules, and scientific hypotheses ### 3. Hardware for AI: GPUs, TPUs, and Quantum Computing - The origin of GPUs and their repurposing for deep learning - Why GPUs are efficient for matrix computations at the core of deep learning - Applications of AI beyond LLMs: video generation, autonomous trajectories, new symmetries, molecule discovery - Discussion of alternative hardware: TPUs, FPGAs, prospects of quantum computing for AI ### 4. Challenges in AI Modeling for Physics and Science - Discrepancies between discrete computer models and continuous physical phenomena - Brian’s struggle with simulating Mercury’s orbit and general relativity - Can AI suggest genuinely novel physical laws, not just optimize or replicate existing knowledge? - Separation between AI architecture/hardware and algorithm/model design ### 5. AI in Scientific Hypothesis Generation & Discovery - How AI models can generate hypotheses and discover fundamental principles (e.g., Lorentz symmetry from LHC data) - The scientific method as an iterative process, and how AI can accelerate every stage except experimental execution, which often still needs humans ### 6. AI in Practical Applications: Traffic Modeling - Rose Yu’s pioneering work on using AI for traffic prediction at Caltech and Google - Deep learning approaches for forecasting on non-Euclidean (graph) networks - Diffusion convolutional neural networks inspired by fluid dynamics and Navier-Stokes equations - Transforming road traffic forecasting: from 10-15 minute prediction limits to 1-hour forecasts, influencing Google Maps and similar systems ### 7. Simulation vs. Imitation in Physical Modeling - Differentiating between fluid simulation methods: Eulerian vs Lagrangian - The role of AI in forecasting vs. traditional numerical solutions to PDEs (partial differential equations) - AI’s speed advantage and accuracy in generating physically plausible outcomes without directly solving the governing equations ### 8. Bottlenecks and Challenges in Scaling AI for Science - Current project: Multimodal Foundation Model for Automatic Hypothesis Generation (GINI) - Data limitations: high cost of curating and generating large-scale, high-dimensional simulation data - Evaluating and supervising scientific AI models, challenges in setting scalable feedback mechanisms ### 9. AI in Pandemic Forecasting and Epidemiology - Parallels between pandemic modeling and traffic forecasting - Use of physics-guided deep learning models and surrogate models for faster, more confident predictions during COVID-19 - Achievements: ranking #1 among 40 teams in national epidemiological forecasting competitions ### 10. Interdisciplinary Research & Motivation - Rose’s approach to solving real-world problems (traffic/congestion, pandemics, climate) - Cross-disciplinary collaboration as a key to impactful research - Advice for listeners: seek out problems you’re passionate about and work with domain experts ### 11. The State and Future of AI: Turing Test, AGI, and Productivity - Has AI passed the Turing test? - Speculation and uncertainty about the timeline for AGI/ASI (Artificial General/Super Intelligence) - AI as an assistant, not a replacement, to human creativity and scientific progress ### 12. Societal Impact, Ethics, and Safety of AI - Guardrails, AI safety research, and the analogy to past technological advances like nuclear power - Concerns about malevolent AI versus the need to advance technology responsibly - Rose’s perspective: continue technological progress with separate but cooperative roles for technologists and policymakers ### 13. AI’s Role in Climate, Fusion, and Societal Challenges - Sustainable development as a unifying theme in Rose’s work: from climate modeling to clean energy and public health - Discussion of the atmosphere as a domain connecting climate change, disease propagation, and nuclear threats - Suggestions for prioritizing safety and innovation in each domain ### 14. The “AI Scientist” Concept - Evolution of the idea: from symmetry detection to equation discovery to hypothesis generation - AI scientists as productivity-boosting assistants, not as replacements for human researchers - Current and prospective tools for scientists: web apps for molecule design, robotics labs for materials science ### 15. The Future of Education in the Age of AI - Will AI/robots replace professors or change the nature of education? - The enduring value of nuanced, interpersonal, and mentorship-based learning - The evolving role of educators: fostering prompting skills, creativity, curiosity, and personal growth ### 16. Reflections, Advice, and Perspectives - Rose’s personal backstory: taking risks, the immigrant experience, evolving research interests - Guidance for young listeners and future scientists: follow your passion, connect with experts, have the courage to pursue “impossible” problems even at personal or professional risk ### 17. Conclusion and Awards - Presentation of the Into the Impossible Medal of Excellence to Rose Yu - Closing remarks on the importance of cross-disciplinary, optimistic, and collaborative approaches to scientific advancement in the AI era --- If you'd like further detail on any segment, or references to specific quotes, let me know!

🎞️ Clipfinder: Quotes, Hooks, & Timestamps

1 / 2

Rose Yu 00:05:20 00:05:32

Why GPUs Matter for AI Progress: "So by leveraging GPUs that was specifically designed to put a lot more weight on processing, matrix multiplication operations is much more efficient, usually like 10 times, if not 100 times more efficient."

Rose Yu 00:09:13 00:09:33

AI and Scientific Discovery: "One of the key example we showed in our work is just by looking at high energy particle physics data from Large Hadron Collider, the model can automatically recognize there is Lorentz symmetry from data without knowing the knowledge of general relativity from Einstein."

Rose Yu 00:16:05 00:16:08

Data-Driven Traffic Simulators: "So sometimes I call this type of model data driven simulator."

Rose Yu 00:17:08 00:17:17

AI in Fluid Dynamics: "So actually for both representations of fluid, there has been AI models that were designed to forecast or to like simulate the movement."

Rose Yu 00:20:10 00:20:24

Scalable AI Evaluation in Science: "it's definitely not very scalable if we just rely on our user to Give our feedback. So we need to come up with a much more scalable way to collect the supervision and to evaluate the model to tell us whether the model is generating the right hypothesis or the wrong hypothesis."

Rose Yu 00:26:44 00:26:58

Viral Topic: Speeding Up Pandemic Forecasts
"We designed physics guided deep learning method that are hybrid of using deep learning inspired by these principles in physics to forecast the progression of pandemic up to four weeks."

Rose Yu 00:28:30 00:28:55

Viral Topic: Turning Personal Frustration into World-Changing Solutions: "Yeah, I think sometimes I think the biggest drive for me is just to make an impact to this world. And you know, when I started working on traffic problem, just because I was stuck in LA all the time in the traffic and I was actually commuting between downtown LA and Pasadena, so I was really frustrated by the situation. I thought I should do something to improve the status quo."

Rose Yu 00:35:00 00:35:06

Viral Topic: Trustworthy AI
Quote: "Using our first principle knowledge as a guardrail for AI scientists is something relatively easy."

Rose Yu 00:40:46 00:41:01

Viral Topic: The Rise of AI Scientists: "Again, the goal is not to replace scientists, but rather have more like a scientific assistant. Right? So these AI scientists will work alongside with our scientists and then they will accelerate each step of the scientific discovery process."

Rose Yu 00:50:59 00:51:09

Viral Topic: Advice to My Younger Self as an Immigrant
Quote: "If I had an opportunity right then I would say, well, why don't you take a little bit more risk and then just trying to do things that you feel you can make an impact on."

👩‍💻 LinkedIn post

1 / 1

🚀 Just had the pleasure of listening to Professor Rose Yu on the "INTO THE IMPOSSIBLE" podcast with Brian Keating, and I am genuinely inspired by how she’s pushing the boundaries of AI and scientific discovery! 👏 From deploying deep learning models that power Google Maps’ traffic predictions to ranking #1 in national pandemic forecasting during COVID-19, Rose Yu is at the forefront of leveraging artificial intelligence for real-world impact. Here are my top 3 takeaways from her fascinating interview: 🔹 **AI as a Scientific Partner, Not a Replacement:** AI may not "feel" emotions or experience “shivers down its spine” like Einstein did, but it *can* generate new hypotheses, discover hidden symmetries, and accelerate research by processing massive datasets—essentially becoming a creative partner to human scientists. 🔹 **Real-World Impact through Data-Driven Models:** Rose’s work with traffic forecasting transformed road planning by shifting from simple statistical models to sophisticated AI-based predictions—essentially allowing systems like Google Maps to accurately forecast traffic an hour ahead. The same approach boosted the response speed for pandemic modeling, reducing scenario simulation times from a week to a single day. 🔹 **Cross-Disciplinary Innovation is Key:** Rose credits her breakthroughs to tackling problems that directly impact her and her community, collaborating with domain experts across physics, climate science, epidemiology, and beyond. Her advice? Pursue projects that ignite your passion, reach out to field experts, and focus on meaningful problems—innovation happens in the overlap! If you’re passionate about AI, scientific discovery, or just looking for a dose of optimism about the future of human-AI collaboration, definitely check out this episode. The future isn’t AI *versus* humans—it’s AI *augmenting* us.✨ #ArtificialIntelligence #ScientificDiscovery #DeepLearning #STEM #Innovation #Collaboration #IntoTheImpossible 🔗 [Listen to the episode and learn more about Rose Yu’s inspiring journey!]

🧵 Tweet thread

1 / 1

🚨 THREAD: Can AI Make Groundbreaking Scientific Discoveries Like Einstein? 🚨 1/ Meet Professor Rose Yu (@UCSD), the computational physics genius whose AI models beat 40 national teams in pandemic forecasting, power Google Maps’ traffic predictions, AND help crack the symmetries of our universe. 🤯 2/ Rose and her team trained AIs on particles smashing together at CERN’s Large Hadron Collider. The AIs automatically found the same deep symmetries in physics that took Einstein *decades* to theorize—from just raw data, no textbooks! 📊➡️🌌 3/ But can an AI ever have a “happy thought” like Einstein’s legendary epiphany about gravity? Or feel shivers down its (virtual) spine? 🤔 “Probably not,” Rose says—but that’s not stopping them from CREATING. 4/ AI is already generating new math theorems, molecules, AND scientific hypotheses. These models process impossible amounts of data, distill our collective knowledge, and then start to imagine beyond it. Creation is happening, even without emotion. 🚀 5/ Physics is unique: experiments test reality, not just logic. Yet Rose’s research shows AIs can uncover physical laws—like Lorentz symmetry—straight from experimental data, missing all the human bias. Wow. 😱 6/ Rose shares how her deep learning models revolutionized traffic forecasting. Before AI: Only 15 minutes of accurate forecasts. After AI: An HOUR! 🚦 Google Maps uses her tech to guide millions daily. 7/ During the pandemic, Rose’s hybrid AI models delivered national-best forecasts—reducing a “what if?” scenario from a WEEK to ONE DAY. Imagine the public health impact. 👩‍🔬🦠 8/ So what’s limiting AI’s scientific superpowers right now? - Need for bigger, more curated datasets 💾 - Costly computing power (GPUs, TPUs, maybe quantum next?) ⚡️ - Creating models trustworthy enough *real* scientists will use! 9/ But Rose is optimistic: “The goal is not to replace scientists—but to have AI science assistants, turbocharging discovery.” Imagine instantly surfacing every relevant result in 3,000 books—while you eat lunch. 10/ On existential risk? Rose is level-headed: Build guardrails, but don’t let fear stall progress. “Any technology has both sides; we need dialogue between policymakers & technologists. Don’t let panic freeze innovation.” 👏 11/ As for the future of education? “AI won’t replace professors. Personal interaction & nuanced teaching matter—but roles will evolve. The true job: bring curiosity and knowledge *out* of students, not just pour facts in.” 12/ Her advice to young scientists? Be bold, take risks, tackle problems *you* care about, and reach out across fields. Impact comes from passion, not following the dogma. 🚀 If you’re excited by the future where AI and humans collaborate to crack the secrets of the cosmos, follow @RoseYu, check out Brian Keating’s “Into the Impossible” podcast, and let’s go beyond the limits of the possible. 🌟 #AI #Physics #Science #Innovation #Podcast #Einstein #DeepLearning #TechForGood 👇 What would YOU ask the first real AI scientist? Tell me below!

🗞️ Newsletter

1 / 1

Subject: 🚀 Can AI Discover the Next Law of Physics? Inside Prof. Rose Yu’s Vision for Augmented Intelligence Hey INTO THE IMPOSSIBLE Podcast Community, We hope you’re ready to have your mind stretched! This week, Dr. Brian Keating welcomes Professor Rose Yu—a trailblazer at the intersection of artificial intelligence, physics, and real-world problem solving—onto the podcast, and the conversation is packed with insights you won't want to miss. **Can AI Feel “Shivers Down Its Spine”?** Professor Yu’s groundbreaking research asks the big questions: Can an AI ever replicate an Einstein “happiest thought” moment? Is it possible for machines to create *new* laws of physics, not just recognize existing ones? The answer might surprise you: AI may lack emotional chills, but it's already generating creative breakthroughs—think new theorems, hypotheses, even molecules. **AI Scientists: Accelerators, Not Replacements** Don’t worry, human curiosity and creativity remain irreplaceable! Prof. Yu envisions a future where AI works as an assistant—“AI Scientists”—standing shoulder-to-shoulder with researchers, sifting massive datasets, suggesting new hypotheses, and streamlining discovery loops in ways never before possible. The aim isn’t to replace scientists, but to amplify what we can ask and how quickly we can find answers. **Real-World Impact: From LA Traffic to Pandemic Forecasting** You’ll hear how Yu’s AI models are doing far more than playing chess or writing bedtime stories. Her work powers Google Maps’ traffic predictions (allowing us to plan that perfect departure time) and topped national charts for pandemic forecasting during COVID-19—with models that cut what-if scenario response times from a week to a single day! **What Holds AI Back—And What’s Next?** Prof. Yu digs into the limitations facing AI science: not just data and computer power (GPU and beyond), but the need for better ways to teach and supervise machines on scientific problems. She also hints at exciting new frontiers—from foundation models fine-tuned for tasks, to collaborations with “robotic labs” that could one day autonomously test new materials and medicines suggested by AI brains. **The Future of Learning and Discovery—Together** Are professors like Brian Keating facing an existential threat from AI? Not quite! Yu believes our uniquely human ability to build relationships, guide, and inspire won’t be replaced. Instead, the future is about partnership between human ingenuity and machine intelligence—a vision that holds promise for scientific leaps and real-world problem solving alike. **AI and Existential Risk: Rose Yu’s Optimistic Perspective** While AI safety matters, Professor Yu cautions against fear-driven paralysis. Her advice? Let technologists innovate, let policy experts focus on safeguards, and foster communication—without halting progress that could benefit humanity and our planet. **Wisdom for Future Innovators** Asked what advice she’d give her younger self (and you, our listeners), Rose says: *Take more risks!* Pursue problems that matter to you, reach across disciplines, and don’t let the doubters win. Impact comes from curiosity, collaboration, and courage. --- 🎧 **Ready to dive even deeper?** [Listen to the full episode now](#) and discover the synergy of AI and human creativity—and what’s coming next in science and society. Want more? Catch Brian’s conversation with Yann LeCun to hear a different perspective on whether AI can *really* understand the world. Don’t forget to like, subscribe, or leave a comment to let us know your favorite moments! Stay curious, The INTO THE IMPOSSIBLE Team P.S. Have a question for Professor Yu or Dr. Keating? Hit reply—we might feature it in an upcoming episode! --- _Transcript attached for your reference—enjoy exploring the full conversation!_

❓ Questions

1 / 1

Absolutely! Here are 10 discussion questions based on this episode of The INTO THE IMPOSSIBLE Podcast with Professor Rose Yu: 1. **Can AI truly achieve creativity in scientific discovery, or is it limited to recombining existing knowledge? How did Rose Yu’s work with discovering symmetries from collider data challenge your thinking about AI’s capabilities?** 2. **Rose Yu distinguishes between AI’s ability to create and its ability to feel emotions. Why does she believe AI can create but cannot feel, and do you agree with her reasoning?** 3. **What are the current limitations and bottlenecks in using AI for scientific research, especially when it comes to processing high-dimensional simulation data, as discussed in the podcast?** 4. **Discuss the ethical and safety concerns mentioned by Brian and Rose related to AI development. Should progress in AI technology be slowed due to fears about misuse, or is open development preferable? Why?** 5. **How does Rose Yu’s interdisciplinary approach—spanning traffic modeling, epidemiology, climate science, and physics—reveal commonalities between fields that seem unrelated? Can you think of another example where methods from one field might benefit another?** 6. **What role do GPUs play in the evolution of AI, and how did they transition from gaming hardware to essential tools for deep learning, according to Rose Yu?** 7. **How does Rose Yu envision the role of “AI scientists” in the future? Do you see AI as more of a collaborator or a potential replacement for human researchers?** 8. **The podcast questions the future of education in a world with advanced AI tools. How might the traditional role of professors change, and what uniquely human aspects of teaching do you think will persist?** 9. **Reflect on Rose Yu’s advice about taking risks and pursuing impactful problems. How might this perspective benefit someone early in their scientific or technical career?** 10. **After hearing Rose Yu’s optimism about human-AI partnerships, where do you personally stand on the spectrum between hope and concern about the rise of artificial intelligence? Why?** Feel free to use or adapt these questions for group discussion, a classroom setting, or even for personal reflection on the episode!

curiosity, value fast, hungry for more

1 / 1

✅ Can AI truly *discover* the fundamental rules of the universe—without a human in the loop? ✅ Host Brian Keating sits down with Prof. Rose Yu, the trailblazing computational physicist, to reveal how AI is already uncovering physical laws from raw data, transforming science, and reshaping everything from pandemics to Google Maps. ✅ On The INTO THE IMPOSSIBLE Podcast, dive into Rose Yu’s mind-bending breakthroughs—from training machines to spot cosmic symmetries at CERN, to revolutionizing traffic forecasting, to launching the new age of the “AI Scientist.” ✅ The future of science isn’t about replacing humans—it’s about supercharging what’s possible, together. Don’t miss this conversation if you want to see how AI and human curiosity are joining forces to push the limits of discovery! 🎧 Listen now and get inspired to imagine what’s next!

Conversation Starters

1 / 1

Absolutely! Here are some engaging conversation starters you can use in your Facebook group to spark discussion about this episode of The INTO THE IMPOSSIBLE Podcast with Professor Rose Yu: 1. **AI & Emotion:** Professor Rose Yu says AI will never feel “shivers down its spine” like Einstein did during his famous revelations. Do you think true creativity or insight requires emotions? Can machines ever really innovate without feelings? 2. **AI Discovered Lorentz Symmetry!** Rose Yu’s team trained AI to find fundamental symmetries in particle physics—without knowing Einstein’s theories. How do you feel about the idea that AI could uncover new physics that humans haven’t even imagined? 3. **AI in the Lab vs. the Real World:** According to Professor Yu, current AI can generate hypotheses from data but still relies on humans (and sometimes robots!) to test them in the real world. What do you think it would take for AI to be truly autonomous in scientific discovery? 4. **From Traffic to Pandemics:** Rose Yu worked on both Google Maps’ traffic prediction and pandemic forecasting. What other “messy,” real-world systems do you think AI should tackle next, and why? 5. **Data-Driven vs. Equation-Driven:** The episode discusses how AI models “simulate” floods, traffic, or pandemics by learning from data rather than solving equations. Do you trust these black-box data-driven simulations more or less than traditional physics-based models? 6. **Hardware Frontiers:** The discussion touches on GPUs, TPUs, and even quantum computing as the “brains” behind AI. What are your thoughts on where the next leap in AI hardware might come from—and how it could change what AI can do? 7. **AI and Academia:** Professor Yu imagines a future where AI is a “scientific assistant,” augmenting rather than replacing scientists and professors. How do you see AI changing education and research careers in the next 10-20 years? 8. **Guardrails or Open Development?** The debate around AI safety comes up—should progress be slowed for fear of risks, or should we separate technology development from regulation? Where do you stand on regulating advanced AI—proactive guardrails or let innovation flourish? 9. **Cross-Disciplinary Inspiration:** Rose Yu found connections between traffic, climate, epidemiology, and AI, based on her own frustrations and experiences. What’s a problem in your daily life you wish AI researchers would take on next? 10. **Podcast Legacy:** Fun fact from the episode: the very word “podcast” traces back to Arthur C. Clarke and the concept of AI! What other parts of pop culture do you think have been shaped by science fiction—and might inspire the next generation of AI breakthroughs? Feel free to personalize or tweak these as needed for your group!

🐦 Business Lesson Tweet Thread

1 / 1

AI just discovered *real* symmetry in physics—without ever reading Einstein. Let’s talk about why that matters and what it means for the future. 1/ Imagine a machine, trained only on raw collider data, independently uncovering Lorentz symmetry—the stuff behind Einstein’s relativity. Not copying, but *discovering*. 2/ Rose Yu’s team at UCSD pulled this off. No built-in physics. No hints. Just data, and the AI figured out the pattern humans took decades to see. 3/ This is bigger than AI winning at chess or folding proteins. It’s AI showing flashes of originality, a nose for pattern, even in the math behind reality. 4/ Why isn’t AI “creative” in the human sense? Because it still can’t feel “shivers down its spine.” It doesn’t dream up new laws with joy or fear. But can it invent? Absolutely. 5/ GPU hardware—originally made for gaming bro-battles—turned out to be perfect for the heavy math behind deep learning. Computer science loves accidents. 6/ But, the real leap isn’t hardware. It’s using data-driven models to draft new scientific hypotheses, spot physical symmetries, synthesize molecules, and cut traffic in LA. 7/ AI is already an *accelerator* for scientific discovery. Faster hypothesis generation, broader literature reviews, smarter experiments. It won’t replace scientists; it’ll make the best ones 100x more effective. 8/ Biggest roadblock? Data quality—not compute. For breakthroughs, curating the right mix of simulation, observation, and expert judgment is everything. Garbage in, garbage out. 9/ Invent the future by finding pain points and reaching out to domain experts. Rose Yu got into traffic modeling because she was stuck in LA. Solve problems that *bug you*; the motivation will be real. 10/ Guardrails? Yes. Fear? Not helpful. The winning teams will build trustworthy, transparent AI and let the speed of discovery outpace the risks. No progress through paranoia. 11/ TL;DR: AI is not about replacing you. It’s about removing the “busywork” from genius, freeing up humans to push further into the unknown. 12/ The only real limit: How bold you are willing to be. Take risks. Pick unsolved problems. Stay curious. And don’t just copy—*discover*. #AI #innovation #science #startups

✏️ Custom Newsletter

1 / 1

**Subject:** 🚀 New Episode: Can AI Be the Next Einstein? | INTO THE IMPOSSIBLE with Prof. Rose Yu --- Hey Impossible Thinkers! We’re thrilled to drop the latest episode of “The INTO THE IMPOSSIBLE Podcast” with a special guest who is truly living up to the show’s name: Dr. Rose Yu, superstar computational physicist from UC San Diego. If AI finding the secrets of the universe, winning pandemic prediction contests, and making your trip down the I-5 faster sounds wild, get ready—this episode is for you. --- 🎧 **In This Episode with Prof. Rose Yu, You’ll Learn:** 1. **The Real Limits of AI “Emotion” and Creativity:** Why Rose believes AI can’t get goosebumps (or “happy thoughts”), but it *can* discover new physical laws, spot symmetries in particle physics, and dream up original scientific hypotheses. 2. **How AI Revolutionized Google Maps Traffic:** Learn how Rose’s deep learning models took traffic prediction from “eh, maybe 15 minutes” to accurate, hour-ahead forecasting—changing the way millions of people get around. 3. **The Science Behind AI-Discovered Symmetries:** Find out how machine learning dug up the same deep laws of nature Einstein spent decades theorizing—just from raw data, no equations required. 4. **Biggest Bottlenecks in AI Science Today:** Hardware, data curation, and the surprising challenge of creating meaningful supervision—plus, what Rose thinks the future of AI-powered scientific breakthroughs looks like. 5. **What It Means to Be a Cross-Disciplinary Maverick:** Rose’s advice for tackling problems big as pandemics and as “everyday” as traffic—plus how she bridges physics, climate science, epidemiology, and AI. --- 🥳 **Fun Fact from the Show:** Did you know the word "podcast" actually traces its origin back to *Arthur C. Clarke’s* 2001: A Space Odyssey? When Hal gets asked to “open the pod bay doors,” it unknowingly set off a chain of events that would name the very thing you’re listening to right now! --- 🎬 **Listen in for inspiration—whether you’re an AI skeptic, a science geek, or just want a traffic-free drive through L.A.** Rose is equal parts optimist and realist—a rare voice saying “AI won’t replace humans, but it *will* massively accelerate discovery (if we let it).” You’ll end up smarter, inspired, and maybe even a little more hopeful for the future. --- ✨ **Ready to go into the impossible with us?** Hit play, leave us a comment with your favorite “Wow, I didn’t know that!” moment, and don’t forget to subscribe so you don’t miss future episodes with guests who are breaking all the boundaries. 🎧 [Tune in here!]([Your Podcast Link]) ⭐ Like, share, and review if you love the show! Till next time, The INTO THE IMPOSSIBLE Team --- P.S. — If you want more mind-expanding AI talk, check out our previous episode with Yann Lecun for a totally different angle on machine intelligence!

🎓 Lessons Learned

1 / 1

Absolutely! Here are 10 key lessons from the episode, each with a concise title and a short description: 1. **AI Lacks True Emotion** AI can generate creative outputs, but it does not experience feelings or instinctive "shivers" like humans. 2. **AI Can Accelerate Discovery** Deep learning enables AI to independently identify scientific symmetries and create hypotheses, speeding up research processes vastly. 3. **Hardware Shapes AI Power** GPUs, built for fast arithmetic, unexpectedly became crucial for deep learning due to their efficiency in matrix operations. 4. **Beyond Games: AI Applications** AI's abilities extend from language models to simulating physical laws, chemistry, music, and even traffic forecasting. 5. **AI Unveils Hidden Symmetries** AI models can detect complex physical patterns, like Lorentz symmetry, directly from data, without preprogrammed knowledge. 6. **Data-Driven Simulations Replace Assumptions** Modern AI simulators learn from massive real-world data, creating more accurate forecasts than models built solely on equations. 7. **Embracing Interdisciplinary Impact** Combining AI with fields like epidemiology and climate science broadens problem-solving approaches and magnifies societal impacts. 8. **Bottlenecks: Data and Evaluation** AI research is limited by simulation data quality, model scalability, and the challenge of finding effective evaluation methods. 9. **AI as Scientific Partner** Instead of replacing humans, AI can become an essential assistant, helping with literature search, hypothesis generation, and experiment planning. 10. **Guardrails and Opportunities** AI's risks must be managed through collaboration and oversight, but fear shouldn't hinder technological progress and the pursuit of global benefits.

10 Surprising and Useful Frameworks and Takeaways

1 / 1

Absolutely! Here are the ten most surprising and useful frameworks and takeaways from the INTO THE IMPOSSIBLE Podcast episode featuring Professor Rose Yu: 1. **AI Can Discover Physical Laws From Data Without Human Bias** - Professor Yu’s team demonstrated that deep learning models, when trained on data from the Large Hadron Collider, could automatically detect physical symmetries like Lorentz invariance—discoveries that originally took human geniuses decades. AI doesn't need to be told the underlying theory; it can reveal them autonomously from raw data. 2. **AI Creativity Is Real, Even Without Emotion** - Though AI lacks feelings like “shivers down the spine,” it can genuinely create. Rose Yu points out examples where AI models generate new mathematical theorems, hypotheses, and even novel molecules. AI’s creativity emerges from vast knowledge absorption and recombination—creativity by information synthesis, not by emotion. 3. **AI as a Data-Driven Simulator Versus Traditional Simulators** - In areas like traffic modeling, Yu’s diffusion convolutional neural networks didn’t simulate traffic via classical equations, but by learning directly from huge troves of real-world data. This approach outperforms equation-based methods, accurately predicting traffic flow up to one hour ahead, and has been adopted by Google Maps. 4. **AI Accelerates Every Step of Scientific Discovery** - Yu describes the entire scientific cycle—observation, hypothesis generation, experimental design, and testing—as fertile ground for AI acceleration. AI’s ability to process tons of data and generate multiple hypotheses empowers scientists to iterate much faster. 5. **Blurring the Boundaries Across Disciplines With Abstraction** - Rose Yu’s work—applying similar AI models to traffic, climate, and pandemics—shows that, fundamentally, many complex systems can be abstracted to similar computational problems. Embracing abstraction enables breakthroughs in fields that seem unrelated at first glance. 6. **Bottlenecks: Curation of Data & Model Evaluation** - One of the main AI limits isn’t hardware, but the effort required to generate high-quality, curated datasets (including multimodal data like text + simulations) and to design effective, scalable evaluation metrics that go beyond narrow benchmarks. 7. **Foundation Models Will Transform Science** - The future is “foundation models” that are pre-trained on massive, cross-domain datasets and then fine-tuned for specific scientific tasks (e.g., causal discovery, molecule design). These models share architectures and can transfer insights across disciplines, fundamentally changing how research is done. 8. **Human-AI Collaboration, Not Replacement** - Yu is optimistic: the role of the AI “scientist” is not to replace human researchers, but to serve as a turbocharged assistant—reading thousands of papers in minutes, generating ideas, and letting humans focus on creativity, synthesis, and judgment. 9. **Importance of AI Trustworthiness and Scientific Guardrails** - Rose Yu argues for embedding first principles and physical constraints into AI models so their predictions don’t violate basic conservation laws. Building trustworthy AI is especially vital for adoption in scientific communities wary of “black box” tools. 10. **Personalized Impact Through Passion-Driven, Cross-Disciplinary Problem Solving** - Her own career strategy is a model: find urgent, real-world problems that bother you, reach out to domain experts, abstract and frame the problem computationally, and use your skills to make a difference. Don’t fear stepping outside your comfort zone—impact comes from pursuing high-pain, high-purpose problems. These frameworks and philosophies—grounded in Rose Yu’s pioneering work—offer a roadmap for anyone interested in harnessing AI for science, crossing-disciplinary boundaries, or navigating the increasingly collaborative future of discovery.

Clip Able

1 / 1

Absolutely! Here are five compelling social media clips from the transcript, each with a suggested title, start/end timestamps, and engaging captions. I’ve ensured each runs for at least 3 minutes and covers thought-provoking moments from the conversation. --- **Clip 1: "Can AI Make Scientific Discoveries Like Einstein?"** **Timestamps:** 00:01:13 – 00:04:41 **Caption:** "Can artificial intelligence ever experience a eureka moment like Einstein—or even come up with new laws of physics? Professor Rose Yu explores the difference between AI’s computational creativity and human intuition, and how machines are already making remarkable theoretical leaps in science." --- **Clip 2: "How AI Shapes Your Daily Commute: The Google Maps Revolution"** **Timestamps:** 00:11:56 – 00:16:08 **Caption:** "Behind your traffic app is deep learning magic. Prof. Rose Yu reveals how her work at Caltech and Google changed traffic forecasting forever—making your ETA smarter, and even predicting collisions. The inside scoop on how data, physics, and AI power your next drive." --- **Clip 3: "AI in the Lab: Can Machines Truly Simulate Nature?"** **Timestamps:** 00:16:08 – 00:19:04 **Caption:** "Will AI ever replace complex scientific equations? Rose Yu explains how deep learning models are not just imitating nature—they’re accelerating simulations from weather to fluid dynamics, revealing the frontier where data-driven AI meets the laws of physics." --- **Clip 4: "Forecasting Pandemics and Traffic with the Same AI"** **Timestamps:** 00:26:04 – 00:29:48 **Caption:** "From crowded highways to global outbreaks, AI uses similar techniques to forecast both. In this clip, Professor Rose Yu reflects on her journey from LA gridlock to pandemic modeling—and shares her advice for tackling big real-world problems across disciplines." --- **Clip 5: "The Future of Science and the AI Scientist"** **Timestamps:** 00:39:46 – 00:43:21 **Caption:** "Imagine an AI partner for every scientist. Rose Yu and Brian Keating dig into the 'AI Scientist' concept—how these digital collaborators could speed up discovery, review thousands of papers, and reshape the entire research process. Are professors ready for their new AI lab partners?" --- Let me know if you’d like shorter clips, or if you want content tailored for a specific platform!

What is Castmagic?