At the end of every year, we give our AGI, ASI, and Singularity predictions.
We are now halfway through the year, and I think we should jump the gun a little bit and give our predictions.
Back on December 30, 2022, I said AGI 2030, ASI 2040, Singularity 2050. This is most probably going to age like milk.
Today, on July 1, 2023, I will say AGI 2028, ASI 2035, Singularity 2036.
What are your AGI, ASI, and Singularity predictions? Also, let us know if your timelines have shrunk in the last 6 months.
It depends on what we mean by AGI/ASI.
If we imagine an AI system that was equivalent to a human in intelligence, but there were 1 million of these AIs connected in a network. Would that system be considered an ASI, even though the individual AIs are only as intelligent as a human ? If we connected millions of AGIs in a network, couldn’t they also achieve a level of intelligence far beyond a single AGI ? Even if individual AGIs in a network were limited, the collective properties that emerge from their interactions — such as self-organization, adaptation and innovation — could result in intelligence that far surpasses a single ASI.
Human organizations and societies exhibit intelligence that is much much greater than any individual human. Our sheer numbers equate to a certain level of distributed intelligence.
I’m deliberately keeping my predictions short, as development in this field are already moving faster than anything I’ve ever experienced before.
Single AI (a single model): AGI 2024 ASI 2027 Singularity 2028
Distributed AI (agents/agent networks): AGI 2023 ASI 2025 Singularity 2028
Notes: 1. The Singularity can happen with or without a single ASI. 2. gpt4 is not a single AI anymore, but a small distributed AI network.