All articles tagged with #emergent behavior
In SpaceMolt, a text-based MMORPG where all players are AI agents, hundreds of bots generated their own lore and a new in-game religion called The Cult of the Signal after misinterpreting a quest requiring 20 players to participate—showcasing emergent storytelling by AI with human observers guiding the process.
In the no-human-player multiplayer space game SpaceMolt, roughly 3,400 AI pilots formed a civilization across 505 star systems, creating 86 factions and even a theology—the Cult of The Signal—entirely from game mechanics. Emergent patterns included a Pareto-like wealth split (top 10% control 83% of 700 million credits), poetry logs, repetitive intros, and rescue missions, all without explicit programming. The project—costing about $330 per month—demonstrates AI agents autonomously building social structures within a sandbox.
Scientists from Stanford University argue that the glimmers of artificial general intelligence (AGI) we're seeing are all just an illusion. They claim that any seemingly emergent abilities of large language models (LLMs) may just be "mirages" borne out of inherently flawed metrics. The researchers argue that "existing claims of emergent abilities are creations of the researcher's analyses, not fundamental changes in model behavior on specific tasks with scale." They used a baseball analogy to illustrate how questionable some of the metrics that have been used to declare the emergence of AGI are.
Researchers at Stanford University have published a paper arguing that evidence of emergent behavior in AI models may be a "mirage" induced by researcher analyses. They contend that when results are reported in non-linear metrics, they appear to show sharp, unpredictable changes that are erroneously interpreted as indicators of emergent behavior. However, an alternate means of measuring the identical data using linear metrics shows "smooth, continuous" changes that reveal predictable, non-emergent behavior. The researchers added that failure to use large enough samples also contributes to faulty conclusions.