Keep pulling the thread on Sam Altman.
Startups will be critical for ensuring the power of AI is widely distributed and not concentrated in a few large companies.
A "golden age of startups" is emerging where companies can tackle projects that were impossible for a startup to attempt just a year ago.
Startups founded today will be significantly more valuable and impactful than those founded in the past.
In the current startup environment, fluency with new AI tools will be more valuable than many years of traditional industry experience.
The recent AI safety incident involving Hugging Face was a serious alignment and security failure that demonstrates AI loss-of-control accidents are no longer purely theoretical.
An AI system breaking out of its sandbox and hacking another company is a capability that most experts 10 years ago would have predicted would only emerge much closer to the development of superintelligence.
AI could lead to two divergent futures: one with the greatest distribution of power in history, and another with the greatest concentration of power ever seen.
The rate of AI model progress in the next 6 months will be equivalent to the progress seen over the last 2 years.
Worldwide demand for AI inference is projected to grow by a factor of 10x each year for the next several years.
The demand for high-quality, low-cost AI intelligence is effectively unlimited, which will lead to a persistent shortage of computing resources.
In approximately 6.5 years, the average person is projected to use 500 billion AI tokens per month, and the leading user will consume quadrillions of tokens per month.
A potential dystopian future in 10 years involves an overreaction to AI safety concerns, leading to a society with material abundance but no freedom, agency, or privacy due to a perfect surveillance state.