AGI needs to outpeform all other computer systems

The notion of Artificial General Intelligence (AGI) hinges on the idea that an intelligent machine should be able to handle any cognitive task that a human can, and do so at a comparable or superior level. To qualify as an AGI, a system should demonstrate flexibility, adaptability, and proficiency across a wide range of domains, rather than excelling narrowly in just a few. One illustrative benchmark in this conversation is chess. While chess by itself is not the ultimate test of general intelligence, it is a well-established challenge that requires strategic thinking, forward planning, pattern recognition, and creative problem-solving.

If an AI cannot play chess at a high level—particularly when there are other specialized chess engines or dedicated computer systems that can routinely outperform even grandmasters—this shortcoming reveals an important point. AGI, by definition, is expected to be at least as competent as the best specialized systems in each domain it attempts. Being outperformed by these narrowly focused, non-general systems suggests a significant limitation. It indicates that the AI’s reasoning processes, adaptability, and learning capabilities fall short of what we would expect from a truly general intelligence.

Moreover, when a supposedly general AI is outperformed by a specialized computer system on a specific task, practical considerations come into play. If humans have access to a specialized chess engine that consistently yields better results, there would be no incentive to use the so-called “general” AI for that task. Instead, people would turn to the specialized system to achieve the best outcome. Over time, this dynamic reinforces the point that the AI is not truly general. A genuinely general intelligence would not need to be sidelined in favor of specialized technology; it would simply excel in chess—or at least match the specialized tool—without any need for human operators to seek more capable alternatives. Instead, by failing to capture the niche that the specialized system has mastered, the AI’s general capabilities are called into question.

Thus, even if this “very smart” AI excels at other tasks, its underperformance in chess—and the resulting impracticality of using it over a specialized engine—undermines the claim that it is a genuine AGI, rather than just another highly specialized form of artificial intelligence.