newsfilter.io
Product Demonstration, Keynote

What will automated firms look like?

  • AGI's primary collective edge is not raw intelligence but digital nature, enabling the infinite copying of workers with intact judgment and tacit knowledge, unlike human employees who are constrained by hiring and training bottlenecks.
  • Capital can be directly converted into compute (electricity, chips, data centers) to sustain populations of billions of digital employees, bypassing the limitations of human physical scaling.
  • Current CEO knowledge is inherently incomplete due to filtered reports and the inability to absorb full context across vast corporate hierarchies.
  • Future "Mega Steve" AI entities will aggregate data from millions of distilled copies, learning from every customer conversation, engineering decision, and market response similar to how Tesla's models learn from millions of driver records.
  • AI firms will operate as rapidly evolving "blobs" of millions of entities that spawn specialized copies and reabsorb knowledge, fundamentally changing social organization to a scale comparable to the transition from hunter-gatherer tribes to joint-stock corporations.
  • AI instances will communicate directly through latent representations, allowing models to merge and propagate ideas across the organization with perfect fidelity, eliminating the biological friction of human social learning.
  • Unlike human societies where knowledge transfer takes years or decades, AI firms can preserve, share, and apply all tacit knowledge instantly across millions of copies.
  • Historical data indicates population size is the primary driver of idea generation; AI firms with massive populations will accelerate cultural evolution and innovation through deliberate experiments, lucky mistakes, and de novo inventions.
  • The cost of high-level roles will be determined solely by the amount of inference compute consumed, shifting scarcity from human talent to computational resources.
  • Companies may justify massive compute expenditures (e.g., $100 billion annually for "Mega Steve") to enable millions of subjective hours of strategic planning, Monte Carlo simulations, and exhaustive scenario analysis.
  • AI firms will optimize for specific valuable abilities (e.g., engineering talent or researchers) by generating marginal copies at penny costs rather than relying on rare human skill distributions.
  • Strategic decisions will utilize massive compute to simulate market dynamics and evaluate thousands of alternative strategies to maximize the value of insights (e.g., simulating FTC responses to major acquisitions).
  • The video production was entirely generated using Google's Veo2 model, demonstrating the capability to produce photorealistic and complex visual narratives without traditional human filming crews.
  • The production utilized Veo2 to generate specific concepts like an FPV drone fly-through of an anthill representing an AGI hive mind, which were stitched together into the final cut.
  • Veo2 is now available for public use via the Gemini app, allowing users to generate video from text prompts.
  • A core distinction between AI and human firms is "evolvability": human corporations cannot effectively replicate themselves or their culture due to reliance on non-copyable human elements.
  • Fully automated companies may achieve a complexity leap analogous to the biological transition from prokaryotic to eukaryotic cells, enabling trillions of coordinated units.
  • Without the ability to replicate perfectly, human firms risk scleroticism and aging, whereas AI firms can clone successful subdivisions instantly.
  • While internal AI planning can be highly efficient, external market feedback remains necessary to prevent internal goals from drifting away from market reality as the firm grows.
  • AI corporations will likely become more software-like, featuring faster feedback loops and perfect replication, potentially shifting the balance between internal planning and market competition.
  • The market provides the necessary measure of real success or failure to anchor AI corporations, preventing them from becoming self-referential entities disconnected from economic reality.