top of page
By Anthony Aguirre
Why should we believe that the timeline to AGI is short?
Here are some of the reasons why:
Massive Investments
Unprecedented resources are being applied to AI development, with the US and EU committing hundreds of billions of dollars to the effort. Goldman Sachs predicts that the world’s largest tech companies and utilities will invest a trillion dollars in AI development over the coming years - many times more than the total cost of the first moon landing.
“Scaling Laws”
AI engineers have observed a trend, similar to “Moore’s Law”, that AI models become more capable as more compute is used to train and operate them. So long as this holds true, we can improve AI capabilities just by adding more computing power. Corporations are poised to increase AI development computing power by orders of magnitude over the coming years.
Insider Signaling
Leaders of AI corporations, who have access to cutting-edge systems, believe “AGI” can be developed within a few years. This is reflected in their public statements, decisions and business strategies.
Expert Predictors
Prediction markets and other platforms where users stake money or reputation on accurate predictions, and which have an excellent track record of past prediction on AI and other technologies, currently indicate that there is a 50% chance we reach “AGI” within 5 years.
AI Bootstrapping
AI systems are increasingly contributing to their own development. The ‘RE-Bench’ benchmark was recently created to measure the ability of AIs to do AI research and development. It found that some of today’s AI models already beat human experts at large, open-ended tasks related to AI engineering. They were also often faster and less costly. Once they become as competent as human AI researchers, they will likely improve at a runaway rate.
bottom of page