The race to AGI is fascinating because unlike most recent macro technology shifts:

1. Not biased for startups -- Both incumbents and startups are at the frontier; unclear if disruptive or sustaining innovation

2. Top talent heavy -- The top 100 people (top 10?) are differentiated talent; feels like the NBA; other technology shifts are more market tailwind + execution

3. No network effect -- It seems like the core technology (foundational models) are not platforms / network effects (and are being rapidly commoditized by open source); the stronger moat is brand recognition and paid customers / subscriptions; Nvidia may be the exception

It might be the wrong race. Instead of AGI>ASI, Collective Intelligence models could solve most if not all alignment issues, use a ton less energy, and be a shortcut to SuperIntelligence mediated through humanity rather than a big separate unpredictable brain.

A sandboxed Human/AI Intelligence cognitive DAO (or better yet, many such DAOs operating as a DAO of DAOs) could, theoretically, teach us how to train neural nets to solve problems more effectively. I think. Still sorting out a mess of ideas and experimental data, but it's certainly possible.