DeepMind researchers outline four paths from AGI to superintelligence

Nov 20, 2026 08:36:22

Google DeepMind research scientist Tim Genewein published an article at the DeepMind Institute, exploring potential development paths following Artificial General Intelligence (AGI). The article describes Super Artificial Intelligence (ASI) as systems that can consistently surpass large groups of human experts on a wide range of human concern tasks and proposes four technical paths from AGI to ASI: expanding model and data scale, emergence of paradigm shifts, enhancing agent collaboration capabilities, and recursive self-improvement.

The article also discusses the impact of hardware investment, algorithm efficiency, and computational power growth on the scale of AI systems and the cost of cognitive tasks, noting that model and data expansion may face challenges such as high-quality data supply, diminishing returns to scale, and limitations on resources like electricity and chips. The author states that assessing AI capability progress requires new benchmarks and modeling methods, with participation from cutting-edge laboratories and institutions at all levels. The article is based on the technical report "From AGI to ASI," co-authored by Tim Genewein and others; the DeepMind Institute clarifies that its article should not be viewed as Google's official stance.

Recent Fundraising

More
Nov 19
$10MNov 19
$1.7MNov 18

New Tokens

More
Dec 31
Dec 31
SSaturnSTRN
Dec 31

Latest Updates on 𝕏

More
Nov 19
SSolanaFollowedcabals
Nov 19
Nov 19