Shelly Palmer: When agents swarm

Think About This: If you work in multi-agent systems, distributed systems, organizational science, economics, AI safety, complex systems, or a related field, I would be grateful for your scrutiny. Greetings from Candlewood Lake. Yesterday, OpenAI launched GPT-6 Astra, calling it 'the world's most intelligent and aligned model.' The rollout starts with a limited set of organizations and will expand over the coming days to all ChatGPT Plus, Pro, Business, and Enterprise users, as well as the OpenAI API, Microsoft Azure, and AWS Bedrock. At a media briefing, OpenAI president Greg Brockman closed by saying, 'Welcome to the AGI era.' For those of you who care about benchmarks, OpenAI reports that Astra scored 97.6 per cent on FrontierMath Tier 4, 99.9 per cent on ARC-AGI-3 using a provider-adapter harness, and 100 per cent on ExploitBench. For context, ARC Prize reports a 62.7 per cent result using its Standard harness, while OpenAI's newer, contamination-resistant ExploitBench version produced a 39 per cent score. ARC Prize Foundation's Greg Kamradt said Astra surpassed the human action-efficiency baseline on 96 per cent of ARC-AGI-3 levels, 'effectively reaching human parity on the benchmark.' ARC Prize nevertheless cautions that this Under what computational conditions does organizational intelligence emerge? The OpenAI/Hugging Face incident gave us a perfect case to study. I have just finished the third version of a working paper, The Emergence of Organizational Intelligence, that explores the question. This is a working paper, not a finished theory. I am trying to understand when a population of bounded agents begins to function as something more like an organization, what mechanisms make that possible, how we might measure it, and what the implications are for enterprises that will increasingly depend on these systems. I am publishing this work in progress because I want help improving the research. I would welcome criticism of the argument, challenges to the terminology, contrary evidence, relevant research I have missed, better experimental designs, and examples that weaken or strengthen the thesis. If you work in multi-agent systems, distributed systems, organizational science, economics, AI safety, complex systems, or a related field, I would be especially grateful for your scrutiny. Read more. -s P.S. I'm going to keynote and host MMA Global's CMO AI Transformation Summit the afternoon of September 17 in NYC. If you're a senior marketer, request your invitation here. About Shelly Palmer