Keep pulling the thread on Sam Altman.
The intelligence of an AI model is approximately equal to the logarithm of the resources, such as compute and data, used to train and run it.
Scaling laws for AI models are accurate over many orders of magnitude, allowing for continuous and predictable gains in intelligence with increased investment in resources.
The cost to use a given level of AI capability is decreasing by a factor of 10x approximately every 12 months.
The socioeconomic value created by linear increases in AI intelligence grows at a super-exponential rate.
Exponentially increasing investment in AI is expected to continue in the near future.
Future software engineering AI agents will be capable of performing most tasks that a software engineer with a few years of experience at a top company can do, for tasks lasting up to a couple of days.
AGI is expected to accelerate scientific progress to a rate much faster than today, potentially surpassing all other impacts of the technology.
The widespread adoption of AGI will eventually lead to a dramatic fall in the price of many goods.
A likely alternative to individual empowerment in AI is its use by authoritarian governments for mass surveillance and population control.
The development of AGI could disrupt the balance of power between capital and labor, potentially requiring early policy intervention.
By 2035, an individual using AI should be able to command the equivalent intellectual capacity of the entire global population of 2025.
OpenAI expects its partnership with Microsoft to continue for the long term.