Damn, it totally sucks that we have to cap the subscriber number of all lemmy/threadiverse communities to 200 subscribers, otherwise nobody can agree on anything and all the upvotes/downvotes on comments tends towards 50/50 splits and any consensus established disappears.
It sucks that Loomio is strictly capped at 200 users too....
It also sucks that the Nepalese weren't able to use Discord to establish a consensus based political revolution that was effective and specific on details of policy because Discord doesn't allow more than 200 people in a room.
https://niemanreports.org/nepal-gen-z-protests-memes-discord-revolution-oli-one-piece/
Waittttt a minutttteeee.... nevermind sorry this is just shitty science/headline writing that fails at the step of framing the question right.
https://www.science.org/doi/10.1126/sciadv.aea6091
The rise of our species did not stem from superior individual intelligence only. Instead, humanity’s unprecedented success emerged from a unique capacity: the ability to coordinate and cooperate in large groups (1). Our ancestors developed sophisticated mechanisms for collective decision-making, like language and writing, that enabled unprecedented scales of collaboration (2). Current large language models (LLMs) demonstrate remarkable individual capabilities across code generation (3), medical diagnosis (4), legal domain (5), sentiment analysis (6), scientific research (7), and mathematical reasoning (8) among others. In many tasks, they have human or even super-human capabilities (9). Yet, despite recent advances in multiagent frameworks (10), today’s LLMs lack the sophisticated coordination mechanisms that enable human groups to tackle complex collaborative tasks (11). The future of artificial intelligence (AI) may thus depend not on creating ever-more-capable individual models but on developing systems of agents that coordinate effectively at scale. This vision is emerging in frameworks such as AutoGPT (https://github.com/Significant-Gravitas/AutoGPT), Microsoft’s AutoGen (12), and OpenAI SWARM (https://github.com/openai/swarm), which enable meaningful interaction between models rather than simple ensembling (13). Recent advances in on-device LLMs and AI-powered assistants are accelerating this transition, with agents beginning to coordinate tasks and make collective decisions on behalf of users (14–16). Trading bots already demonstrated how agent interactions can produce emergent behaviors like flash crashes (17), while recent releases of Google’s Agent2Agent protocol (https://github.com/google/A2A) and dedicated multiagent libraries like Concordia (18) signal growing industry recognition of this paradigm shift. Just as we study animal behavior without fully understanding neural mechanisms, we need to examine AI agents as social actors whose collective patterns can be systematically analyzed despite opaque internal workings.
throws up on keyboard like fuckkk this is basically saying we should as an axiom approach "AI Agents" as entities that are social and sentient... which is an incredibly unstable and biased axiom to start the process of studying something from....