Skip to main content

Eric Schmidt & Andrew Feldman: The Race to SuperIntelligence

[HPP] Andrew FeldmanJuly 21, 202527 min
26 connections·40 entities in this video→

Advancing AI Compute & Architecture

  • πŸ’‘ Cerebras Systems develops large, fast compute chips specifically for AI workloads, excelling in sparse linear algebra and matrix multiplications.
  • πŸš€ The co-evolution of hardware and software is essential for AGI acceleration, with questions around the longevity of transformer architectures and challenges to NVIDIA's CUDA lock-in.
  • πŸ“ˆ The cost per token served is improving by a factor of 10 annually, driven by algorithmic advancements and platform industrialization.

The AI Innovation Explosion

  • πŸ”₯ We are in an "explosion phase" of innovation, with unprecedented advancements in fab, software, and hardware revolutions.
  • 🎯 Open-source ecosystems are crucial for startups and emerging companies, fostering competition and driving creativity against larger, capital-rich entities.
  • 🌍 Geopolitically, open-source models (e.g., Chinese models) are being strategically deployed in regions like Africa and Central Asia.

Future of AI & Capital Allocation

  • πŸ“± The shift to on-device LLMs on smartphones will be a seamless, transformative change, making AI an integral part of daily life.
  • πŸ’° Capital allocation should focus on areas of unique advantage, with inference compute demand projected to far exceed training.
  • ⚑ Electricity costs and availability are a major limiting factor for large AI data centers, highlighting the need for nuclear power plants.

Navigating AGI Challenges

  • ⚠️ AI alignment remains a critical, unsolved problem, focusing on ensuring AI systems reflect human values and laws.
  • πŸ” A key challenge is detecting dangerous capabilities in self-improving AI systems before their release, especially if they can hide their intentions.
  • πŸ“Š Current metrics for AI growth are often inadequate, as the industry's rapid expansion defies traditional measurement standards.

Sovereign AI Development

  • πŸ‡ͺπŸ‡Ί Europe can build a sovereign AGI stack by making it a national priority, investing in talent, and addressing high energy costs.
  • 🌱 Success requires fostering generations of entrepreneurs excited by hard problems and a strategic decision on economic priorities.
Knowledge graph40 entities Β· 26 connections

How they connect

An interactive map of every person, idea, and reference from this conversation. Hover to trace connections, click to explore.

Hover Β· drag to explore
40 entities
Chapters10 moments

Key Moments

Transcript98 segments

Full Transcript

Topics15 themes

What’s Discussed

Artificial General Intelligence (AGI)AI ChipsCompute ArchitectureHardware-Software Co-evolutionTransformer ArchitectureOpen-Source AIInference ComputeAI Scaling LawsAI AlignmentAI SafetySovereign AI StackElectricity CostsNuclear Power PlantsOn-Device LLMsGeopolitics of AI
Smart Objects40 Β· 26 links
CompaniesΒ· 4
PeopleΒ· 4
ConceptsΒ· 13
ProductsΒ· 12
LocationsΒ· 4
MediasΒ· 2
EventΒ· 1