AI Is Now Discovering Things Humans Cannot

In 2025, Google's AI solved a math problem humans had been stuck on for over three hundred years. It also designed medicines that don't exist in nature, invented materials humans never thought to look for, and figured out the laws of physics on its own, with no textbooks, no equations, and no human help. This is the story of how AI stopped being just a tool, and started making real discoveries on its own. Every claim in this video is pulled from published research by Google DeepMind, MIT, Peking University, and Google AI. What you're about to see is the clearest evidence yet that modern AI systems are uncovering things humans never could have found on their own, not in some far-off future, but right now, in labs around the world. At the same time, this is only the beginning of what may be the most exciting period of discovery in human history. In this video, we cover: – The AI that designed 36 million brand new medicines from scratch – How one model found over 2 million new materials humans never knew could exist – The AI that solved a 300-year-old math problem in just a few hours then started making itself faster – The system that rediscovered the laws of physics with no prior knowledge – What it means when AI starts answering questions we didn't even know to ask – Why the next twenty years may produce more scientific breakthroughs than the last four hundred ⏱ Timestamps 00:00 - Intro 00:40 - AI Inventing New Antibiotics 01:33 - New Materials 04:03 - Retina Scans & Heart Disease 05:15 - Burned Papyrus In Pompeii 06:08 - The kissing Number 06:50 - Alpha Evolve Solving Problems 07:28 - AI Re-Inventing Physics 09:31 - Answering Questions We Did Not Ask 11:27 - Final Thoughts 🔔 Subscribe for more AI documentaries 🛠 Explore 49,000+ AI tools: https://theresanaiforthat.com 📚 RESOURCES & FURTHER READING MIT & AI-DESIGNED ANTIBIOTICS 36 million novel compounds (Cell, Aug 2025): https://news.mit.edu/2025/using-gener... Halicin (Cell, Feb 2020): https://news.mit.edu/2020/artificial-... Abaucin (Nature Chemical Biology, May 2023): https://www.nature.com/articles/s4158... GOOGLE DEEPMINDGNoME (MATERIALS DISCOVERY) Scaling deep learning for materials discovery (Nature, Nov 2023): https://www.nature.com/articles/s4158... Millions of new materials announcement: https://deepmind.google/blog/millions... Science magazine coverage: https://www.science.org/content/artic... GOOGLE AI + VERILY — RETINAL HEART ATTACK PREDICTION Prediction of Cardiovascular Risk Factors from Retinal Fundus Photographs via Deep Learning (Nature Biomedical Engineering, 2018): https://www.healthcareitnews.com/news... World Economic Forum coverage: https://www.weforum.org/stories/2018/... VESUVIUS CHALLENG: DECODING ANCIENT SCROLLS National Geographic: First Vesuvius passages deciphered: https://www.nationalgeographic.com/pr... CNN — Herculaneum scroll decoded: https://www.cnn.com/2024/02/07/world/... Smithsonian Magazine: Three students decipher first passages: https://www.smithsonianmag.com/smart-... Popular Science: https://www.popsci.com/technology/ves... GOOGLE DEEPMIND: ALPHAEVOLVE AlphaEvolve announcement (May 2025): https://deepmind.google/blog/alphaevo... AlphaEvolve paper (arXiv): https://arxiv.org/abs/2506.13131 IEEE Spectrum coverage: https://spectrum.ieee.org/deepmind-al... Newton & the Kissing Problem (Plus Magazine): https://plus.maths.org/content/newton... PEKING UNIVERSITY: AI-NEWTON AI-Newton paper (arXiv, April 2025): https://arxiv.org/abs/2504.01538 Nature News coverage: https://www.nature.com/articles/d4158... Peking University announcement: https://newsen.pku.edu.cn/news_events... AI IN DRUG REPURPOSING & DISCOVERY AlphaFold2 implications for drug discovery (ScienceDirect): https://www.sciencedirect.com/science... AlphaFold2 in biomedical research (Frontiers in Molecular Biosciences): https://www.frontiersin.org/journals/...