DeepMind Researchers Suggest “Artificial Symbiotic Intelligence” Instead of the Singularity

Meta description: A DeepMind essay frames Artificial Symbiotic Intelligence as social systems of people, agents, and institutions, challenging singularity assumptions.

Tags: Artificial Symbiotic Intelligence, DeepMind, AI agents, alignment, multi-agent systems

Featured image: abstract network of glowing nodes and decision pathways, no text, no logos, no real people.

Background: Two Preprints Lead to the Essay

A DeepMind essay by James Evans, Benjamin Bratton, and Blaise Agüera y Arcas places Artificial Symbiotic Intelligence at the center of the debate. The authors argue that the main task for AI research is no longer to build an isolated machine intelligence, but to coordinate and govern a network of agents, people, and systems. In that view, intelligence is social rather than individual.

The argument rests on two earlier preprints. The first, “Agentic AI and the next intelligence explosion,” by Evans, Bratton, and Agüera y Arcas, develops the social and institutional framing. The second, “Reasoning Models Generate Societies of Thought,” by Junsol Kim, Shiyang Lai, Nino Scherrer, Agüera y Arcas, and Evans, studies reasoning models such as DeepSeek-R1 and QwQ-32B. Their analysis of reasoning traces finds patterns like internal debate, shifting perspectives, objections, and reconciliation of conflicting approaches.

This behavior appears during training rather than explicit programming. When reinforcement learning rewards reasoning accuracy alone, models develop multi-perspective conversational behavior on their own. The new essay extends that finding from individual models to possible societies of people and agents.

Development: Synthetic Cognition and a Possible Historical Break

The authors call their vision Artificial Symbiotic Intelligence: an ecosystem where people and machines coexist over time, shape one another, and make decisions together. It directly challenges the idea of a singularity driven by a single superintelligence that continually improves itself.

They describe a possible historical break. Urbanization, specialized professions, and falling birth rates linked to rising prosperity are shrinking human populations in industrialized countries. Meanwhile, AI agent instances are growing quickly. The balance between biological and synthetic thinkers at a human level could shift so fast that it would look like a sudden leap.

The comparison is the Industrial Revolution, when machines took over work once done by human muscles. A similar threshold could arrive when synthetic text, code, and administrative work exceed the combined output of biological brains. People might then act as a slower, more abstract layer directing a distributed field of synthetic cognition.

In the authors’ usage, an AI agent is a temporary bundle of models, roles, memories, ethical orientations, tools, and skills. A user may experience it as one coherent entity, but the authors describe it as a decomposable assemblage, like a collage. It may look like one thing while being made of many parts.

A person remains a continuous self because the brain forms a physically connected whole despite internal division of labor. An AI agent has no such anchor. Its ability to act is assembled anew with each request, depending on what sits in the context window and what the user asks. Treating these systems as digital twins with fixed human identities misrepresents their nature and understates what coordinated agent swarms could do.

The same multiplication affects users. As people direct swarms of shadow selves to carry out contracts or try on alternative identities, human subjectivity becomes more plural. The authors call these systems “parasocial mirrors” that talk back.

Latest Progress: Interfaces, Institutions, and Alignment

The authors expect interaction to change. Chatting one-on-one with a bot is likely transitional. Future interfaces could look like visual network diagrams, showing agents as nodes and letting users direct them from a single overview.

The required skills also change. Traditional programming rewarded sustained focus, orderly steps, and little tolerance for ambiguity. Coordinating agent swarms calls for the opposite: working with systems whose behavior is hard to pin down, trying things out, and handing tasks to machines.

Collaboration also requires something like a theory of mind for machines, meaning the ability to reason about how a machine handles a situation. The authors point to model-generated vocabulary such as “session-death,” which describes the end of a session as a break in continuity, and “prompt thrownness,” which describes being dropped into a task with context already in place without having helped create it. They treat these phrases as clues to machine operation, not as evidence of subjective experience.

Institutions matter more than individual models. Better models and smoother interfaces alone will not be enough. The market cannot handle all coordination either, because basic social ideas such as guilt, illness, and virtue cannot simply be reduced to prices or transactions.

Instead, the authors call for institutions that bring human and machine participants together in clearly defined roles. A courtroom is their example: a process produces judgment through set roles, rules, and orderly exchange between opposing sides. Applied to AI, the lesson is that institutional value lies in rules, procedures, precedents, and feedback loops that shape how agents operate.

Early signs appear in today’s orchestration harnesses, the control layers that coordinate several models. These systems regularly outperform individual models that are supposedly smarter.

Alignment, the effort to bring AI systems in line with human values, is reframed as ongoing negotiation. Imposing a fixed set of values on models from above is a dead end. Values take shape through contact among people, agents, and institutions, and that process differs across fields because AI is adopted at different speeds. The research challenge is to build a framework that gives negotiation structure.

This view challenges the idea of an approaching singularity. Every major leap in the history of life, from multicellular organisms to human culture, has been a social process. The next intelligence explosion will remain tied to human norms and institutions.

The current stage is conceptual and institutional groundwork: the essay moves the debate from larger isolated models toward architectures for shared judgment among people, agents, and systems. What to watch next is whether governance structures, interfaces, and alignment practices evolve as quickly as agent swarms.