Episode
The Path to AGI: Grok 3, Quantum Breakthroughs, and Humanoid Robots
- Published
- Feb 27, 2025
- Duration seconds
- 6206
- Processing state
processed
Actions
POST https://stenobird.com/v1/public/podcasts/generative-ai-meetup/episodes/the-path-to-agi-grok-3-quantum-breakthroughs-and-humanoid-robots/transcription-requests
Idempotently request low-priority transcript generation for this episode.GET https://stenobird.com/podcast/generative-ai-meetup/the-path-to-agi-grok-3-quantum-breakthroughs-and-humanoid-robots.md
Read the agent-friendly Markdown representation of this episode resource.
Summary
Youtube: https://www.youtube.com/@GenerativeAIMeetup https://www.anthropic.com/news/the-anthropic-economic-indexhttps://x.com/xai/status/1891699715298730482 - Grok 3http://x.com/karpathy/status/1891720635363254772https://news.microsoft.com/source/features/innovation/microsofts-majorana-1-chip-carves-new-path-for-quantum-computing/https://en.wikipedia.org/wiki/Majorana_fermionhttps://www.youtube.com/watch?v=88vEsL5tgDIGrok 3-Elon Musk new model- Seems to be a very brute-force approach (which is working)- Definitely a state of the art model- 40 dollars a month now- https://techcrunch.com/2025/02/19/x-doubles-its-premium-plan-prices-after-xai-releases-grok-3/- 100,000 GPUsMajorana- Uses a particle called the Majorana- https://www.howtopronounce.com/ettore-majorana- https://en.wikipedia.org/wiki/Ettore_Majorana- Ettore Majorana disappeared soon after theorizing this particle in 1937 he disappeared in 1938 although thought to be alive- A majorana is a fermion that has its own antiparticle - this was originally theorized in 1937- The particle has its own charge... and own antiparticle- it has a new topological state... which is a new state of matter- Has has 8 quibits- Is designed to scale to a million quibits- To break bitcoin you might need ~1500 to 5000 logical quibits- A quibit can be used to represent potentially infinite statesFigure AI - Helix- https://www.figure.ai/news/helix- Vision-Language-Action (VLA)- Full-upper-body control- Multi Robot collaboration- Pick up anything- Single neural network to build this, no specific fine tuning- https://www.figure.ai/news/helix- Has a slow System 2 and a fast system 1 neural network- system 2 -- 7B- system 1 -- 80M Get ready to explore the frontiers of technology in this exciting episode! We unpack the latest breakthroughs dri…