OpenAI and Microsoft are dueling over AGI. These real-world tests will prove when AI is really better than humans.

The ongoing debate over Artificial General Intelligence (AGI) is heating up, particularly as it pertains to the influential partnership between OpenAI and Microsoft. With the stakes high, the conversation centers around when AI will achieve capabilities that can rival human intelligence. Currently, OpenAI is in a unique position where it must deliver on the promise of AGI to retain its lucrative partnership with Microsoft. This relationship grants Microsoft substantial technological and financial advantages, including a significant share of OpenAI's revenue, amounting to billions of dollars. Some observers suggest that this financial incentive may be why OpenAI's CEO, Sam Altman, frequently emphasizes the organization's proximity to achieving AGI. Conversely, many experts in the AI community remain skeptical about the AGI narrative, arguing that it may be overhyped or not as consequential as it's being portrayed. Dario Amodei, CEO of Anthropic and a strong advocate for AI, often refrains from discussing AGI, reflecting a broader ambivalence among some professionals in the field. In stark contrast, Microsoft's CEO Satya Nadella has a vested interest in the AGI discourse. Given the financial implications tied to OpenAI's AGI milestones, Nadella's skepticism about the timeline for achieving AGI is hardly surprising. He has described the pursuit of AGI as "benchmark hacking," a term that critiques the tendency of AI labs to optimize their models for specific industry benchmarks rather than practical, real-world applications. OpenAI defines AGI as "highly autonomous systems that outperform humans at most economically valuable work," but this definition is subject to interpretation. Achieving AGI involves creating systems that can perform a wide array of tasks better than humans, a feat that remains elusive. In light of these developments, I have devised several practical tests aimed at assessing whether we have indeed reached AGI. These tests focus on everyday tasks that should be seamlessly managed by an AGI system. While they may seem trivial, they highlight the shortcomings of current AI technology, which still struggles with complex real-world scenarios. As OpenAI and Microsoft navigate their partnership dynamics, experts may be called upon to evaluate whether OpenAI has indeed realized AGI, according to the terms of their contract. The conversation around AGI is enriched by insights from AI research scientists, such as Konstantin Mishchenko from Meta, who argue that while large language models (LLMs) can mimic intelligence, they lack the essential algorithms for experiential learning. This raises questions about the true proximity of AGI and whether it is achievable with current technology. As this intriguing saga unfolds, the tech world watches closely, eager to see how the competition between these two giants will influence the future of artificial intelligence.