DeepMind maps the road from AGI to superintelligence
According to Google DeepMind's Tim Genewein, AGI is near and four compounding paths could lead to superintelligence, so curbing just one won't stop it.

DeepMind maps the road from AGI to superintelligence
According to Google DeepMind's Tim Genewein, AGI is near and four compounding paths could lead to superintelligence, so curbing just one won't stop it.
institute.deepmind β€’
Revision history

2 recorded changes

Want your article here?

Promote with Leviathan News

Google DeepMind's Tim Genewein writes that global effective compute stock grows at least 10x per year, a factor of 10 billion over ten years, driven by hardware, algorithmic efficiency and investment. A cognitive workload costing $1 billion today would cost $1000 in ten years; an Epoch study finds benchmark prices fell more than 13x per year since 2023. He maps four pathways from AGI to ASI: scaling models and data, technological paradigm shifts, agent coordination, and cognitive industrialization. Scaling laws are extrapolations, he notes, with data exhaustion, diminishing returns and physical limits as bottlenecks. AI agent populations might scale more than a thousand-fold by 2030.

AI summary of the linked source β€” not editorial content.

Explore the topic

Loading related coverage…

Comments

Loading comments…