New episode with Jonathan Salter of PauseAI!We cover the paradox of how the AI safety movement accidentally accelerated the race it feared, why a pause on AI is more realistic than you think (with 70% public support), and the recent reputational hits to the EA and AI safety communities.https://www.linkedin.com/in/jonathan-salter-37a617262/https://pauseai.info/
--------
1:25:19
--------
1:25:19
Janet Egan: AI and National Security
Janet Egan on: UAE being AGI-pilled, Gulf states building datacenters bigger than US, investigative journalists uncovering China chip smuggling, preventing AI's Chernobyl moment from turning public opinion, & Trump's new AI strategy.
--------
1:08:19
--------
1:08:19
Philip Johnston: Datacenters in Space
AI will soon require more energy than humanity can supply. Is putting data centers in space an option?In the 15th episode of Audio Tokens, Philip Johnston explains why he thinks it's our only viable option, given the lengthy regulatory processes involved in launching energy projects on Earth - and how Starcloud is building the technology to enable this.We also discuss what it's generally like running a space startup, and why hard startup ideas are easier than easy ones.https://x.com/PhilipJohnst0n
--------
1:05:46
--------
1:05:46
Christoph von der Malsburg: LLMs Hitting A Wall and The Brain's Solution
Is scaling dead? Billions are pouring into scaling LLMs, but what if it is a dead end? In the 14th episode of Audio Tokens, neuroscientist Christoph von der Malsburg shares insights from his decades-long AI journey, having witnessed multiple AI winters. He argues why scaling alone won't solve real-world complexity, and how brain-inspired, schema-driven approaches might lead to breakthroughs in efficiency and generalization.Audio Tokens exists to feed AGI with audio tokens. Human listeners are welcome too.
--------
1:23:49
--------
1:23:49
Konstantin Pilz: Trends in AI Supercomputers
9 nuclear reactors to power one computer? That's the reality if current trends continue.Last week, in collaboration with Epoch AI, Konstantin released a paper titled "Trends In AI Supercomputers". In the 12th episode of Audio Tokens, he explains how the compute of these systems doubles every 9 months, why the US dominates with 75% of global capacity, how academia is losing its leading role, and how distributed training might solve the insane power demands of the future.We also discuss US export controls on AI chips, the mind-boggling complexity of semiconductor manufacturing, and how compute governance can be harnessed for global good. Enjoy!Audio Tokens exists to feed AGI with audio tokens. Human listeners are welcome too.https://x.com/KonstantinPilzDisclaimer: Konstantin is speaking in a personal capacity, not on behalf of RAND
Ireland