OpenAI's Joshua Achiam: Did We Already Reach AGI?

This discussion from OpenAI’s Joshua Achiam presents a critical re-evaluation of what constitutes Artificial General Intelligence, suggesting we may be operating within an AGI era without acknowledging it, thereby challenging your fundamental assumptions about AI's current capabilities and future trajectory. The core argument posits that AI's rapid advancements in areas like cybersecurity, coupled with its increasing ability to self-improve and discover novel solutions, are already demonstrating hallmarks of general intelligence, even if these manifestations don't fit our conventional, often sci-fi-inspired, definitions of AGI. This perspective underscores a gradual, pervasive integration of sophisticated AI rather than a singular, dramatic AGI "arrival," forcing us to consider the implications of living with AI that can outpace human problem-solving in complex domains. For working developers, founders, and operators in Zimbabwe, understanding this subtle shift can unlock significant strategic advantages. Consider a logistics startup in Bulawayo dealing with the intricate challenges of fuel delivery routes and unpredictable road conditions; if their AI-powered optimisation tool, initially designed for traffic flow, begins autonomously identifying novel ways to reroute entire fleets based on real-time data feeds about local unrest or unexpected road closures, that’s not just optimisation—it's a system exhibiting adaptive general problem-solving. Similarly, a small e-commerce shop owner in Harare, struggling with sophisticated phishing attempts, might find their off-the-shelf security software, informed by new generative models, not only blocking known threats but proactively identifying and nullifying entirely new attack vectors, demonstrating an advanced form of intelligent defense. An independent SaaS founder building an ed-tech platform for schools across Mutare could leverage this understanding to design features that allow AI to autonomously generate tailored learning paths and even suggest syllabus adjustments based on real-time student performance and regional curriculum updates, moving beyond rule-based systems to a more fluid, intelligent educational assistant. The practical implication here is to shift from viewing AI as merely a tool for automation or specific task execution to recognizing its potential for emergent, general problem-solving. To capitalize on this, take one recurring, complex operational challenge in your business—perhaps supply chain disruptions, customer support anomalies, or unexpected system failures. Instead of designing a fixed, rule-based AI solution, explore how you might deploy a modern large language model (LLM) or a specialized AI agent to observe the problem, access relevant data, and then prompt it to propose novel solutions or even identify entirely new problems you hadn't considered. The goal is to move beyond simply automating existing processes and towards leveraging AI for discovery and autonomous, intelligent adaptation.