Deepmind researchers propose Artificial Symbiotic Intelligence as alternative to singularity

The essay builds on earlier preprints by the same authors and colleagues, showing that reasoning models such as DeepSeek‑R1 and QwQ‑32B develop debate‑like patterns during training when rewarded only for reasoning accuracy. The authors extend this observation to propose that future AI systems will be assembled from many temporary agents—bundles of models, tools and memories—coordinated through new interfaces and institutional frameworks. They stress that governance, alignment and skill development must focus on coordinating swarms of agents and designing institutions, like courts, that define roles, rules and feedback loops for human‑machine collaboration.

If the shift toward networked AI agents occurs, the balance between biological and synthetic thinkers could change rapidly, resembling past industrial revolutions. The authors suggest that research, regulation and market strategies should address this collective intelligence rather than chasing ever larger isolated models.