Resultados
17 artículosOrdenar
1:28:08ENThe same H100 costs 3x at AWS, and a four-month-old lab is catching genes the top sequencers missedPresenta: Nathan LabenzA Cognitive Revolution weekly briefing on a possible U.S.–China AI hotline, why hyperscalers charge two to three times as much for identical chips, co-writing a novel with frontier models, and an AI loop that re-diagnosed a child a premier prenatal lab had cleared.
1:37:43ENCisco wants two extra network layers so agents can share intentPresenta: Nathan LabenzVijoy Pandey of Outshift by Cisco argues that superintelligence has to scale out through networks of agents, not just up through bigger models.
2:12:31ENMost Caffeine users have no idea they are on a blockchainPresenta: Nathan LabenzDFINITY's Dominic Williams tells Nathan Labenz why a tamper-proof, unstoppable cloud is the right backend for AI that writes software from a chat box.
2:06:09ENOne ChatGPT query is a second in the microwavePresenta: Nathan LabenzAndy Masley walks Nathan Labenz through the energy and water math of AI, from a single prompt to an 80-gigawatt buildout.
1:32:47ENA transformer co-author wants a $160 million model no company ownsPresenta: Nathan LabenzIllia Polosukhin tells Nathan Labenz how NEAR turned a payments headache into a stack for private inference, shared-revenue training, and AI agents that still answer to a court.
3:17:35ENTen Thousand Comments and 76 Spare Gigawatts: Four Months Writing America's AI Action PlanPresenta: Nathan LabenzDean W. Ball spent four months as the White House's senior AI adviser, then left as the Action Plan landed — and explains how a to-do list got written inside a building that cannot run ChatGPT.
1:45:00ENAmerica Holds 45% of the World's Data Centers. Everyone Else Must Build, Buy, or AllyPresenta: Nathan LabenzAnjney Midha of a16z maps what “sovereign AI” actually means for companies and for countries that will never own the talent, data, and compute frontier.
2:10:07ENA $30,000 Operation Could Knock a $10 Billion Superintelligence Data Center Offline for a YearPresenta: Nathan LabenzGladstone AI’s Jeremy and Edouard Harris inventory what an American Manhattan Project for superintelligence would actually have to secure — and what it would cost in concrete, supply chains, and civil liberties.
54:55ENNew Jersey Cut Phone Waits from 40 Minutes to 90 SecondsPresenta: Nathan LabenzGovernor Phil Murphy on the state AI assistant, a Princeton hub, and why he wants the best catcher's mitt in the country when superhuman systems arrive.
2:16:17ENA 10B Model Trained Across Three Continents — and an H200 for $1.49 an HourPresenta: Nathan LabenzPrime Intellect's founders lay out a four-step plan to turn spare global GPUs into collectively owned intelligence, and argue that reinforcement learning just made their job easier.
2:11:15ENA 50% Chance of Extinction — and a $50 GPT-4 for EveryonePresenta: Nathan LabenzEmad Mostaque tells Nathan Labenz why he puts p(doom) at even odds, why a five-watt model changes the energy debate, and how the Intelligent Internet would fund universal basic AI with a Bitcoin-like token.
31:04ENThe Risk Is Not China Copying. It Is America Refusing to Use the ToolsPresenta: Nathan LabenzAt the 2024 Hill & Valley Forum, Alex Wang, Jack Clark, and Andrew Ng tell Senator Cory Booker that Chinese labs are inventing, open-source models are already strong, and the Pentagon’s real AI failure is under-adoption.
1:52:18ENTwo Weeks, Thirty Compiles: How Adam Majmudar Wrote a GPU From ScratchPresenta: Nathan LabenzA Cognitive Revolution walkthrough of TinyGPU, from SIMD kernels and memory latency to an open-source chip flow that runs on a laptop.
1:42:17ENThe Cheapest Inference Provider Is the Most DangerousPresenta: Nathan LabenzFireworks AI cofounders Lin Qiao and Dmytro Ivchenko explain why subsidized GPU bills vanish, how LoRA serving can cut cost a hundredfold, and why OpenAI is the rival that actually matters.
1:19:52EN1,000 H100s Take 115 Days; 22,000 Take FivePresenta: Nathan LabenzNathan Labenz and Trey Kollmer walk through cluster math that turns GPT-4-scale training into a weekly experiment, then four papers on search-augmented models, recursive code improvers, spatial maps inside Llama 2, and what networks do when they have no idea.
1:01:26ENAlexNet Used Two GPUs. GPT-4 Took Twenty Thousand. The Shortage Will Be Back.Presenta: Nathan LabenzRun:ai CTO Ronen Dar on filling idle accelerators, why H100s can be cheaper per experiment, and why Llama 2 undercuts compute governance.
1:20:49ENA Wafer-Scale Chip So AI Stops Waiting on AmdahlPresenta: Nathan LabenzAndrew Feldman, CEO of Cerebras, tells Nathan Labenz why one 2.5-trillion-transistor die, sparse linear algebra, and a 16-machine supercomputer are a better shape than a GPU cluster.









