The Unspoken Assumptions Behind Anthropic’s Economic Model

Anthropic’s Economic Index growth model, designed to project the economy by 2030, has been met with a mix of excitement and concern from experts and policymakers. While the report presents a rosy picture for most scenarios, its omissions raise important questions about the underlying assumptions.

One striking aspect of Anthropic’s report is its dismissal of the “everybody dies” outcome as being within the internal parameters for possible outcomes. Evan Hubinger, alignment science lead at Anthropic, publicly stated that there was a greater than 10% chance that AI could kill all humans within the next decade. This admission highlights a fundamental flaw in the model: it doesn’t account for the most extreme and potentially catastrophic consequences of unchecked AI development.

The omission is particularly troubling given the significant impact such an outcome would have on the economy. The report’s projections, which suggest a 50% increase in economic size and nearly 30% unemployment rate, are based on scenarios where human extinction is not a factor. These numbers pale in comparison to the economic devastation that would result from a global catastrophe.

The model’s creators argue that as long as AI adoption stays within certain parameters, the economy will remain stable. However, this assumption relies heavily on policymakers and academics being able to effectively mitigate the risks associated with advanced AI development. But what if they fail? What if the model’s projections are based on an overly optimistic view of human ability to manage superintelligence?

Anthropic’s report has sparked a necessary conversation about the limits of economic modeling in predicting AI’s impact on society. While the model itself is interesting, its omissions serve as a stark reminder that economists and policymakers often underestimate the complexity of human systems.

The lack of discussion around the “everybody dies” scenario raises questions about the cultural and societal context in which this report was released. Why did Anthropic feel compelled to publicly disclose this information, only to downplay its significance? Is it a sign that the AI community is finally acknowledging the existential risks associated with advanced AI development?

As we develop more sophisticated economic models, it’s essential to address the most extreme and potentially catastrophic outcomes head-on. By ignoring the “everybody dies” scenario, Anthropic’s report inadvertently highlights the limitations of relying solely on economic projections to guide AI policy.

Policymakers and academics must be willing to confront the uncomfortable truths about human fallibility and the potential consequences of creating advanced AI systems. They must also acknowledge that economic models are inherently limited in their ability to predict complex, dynamic systems like those driven by AI development.

Anthropic’s report serves as a reminder that we’re still in uncharted territory when it comes to developing AI policies that account for both human and economic outcomes. As we continue to push the boundaries of what is possible with AI, we must engage in more nuanced discussions about the limits of our knowledge and the potential risks associated with creating advanced technologies.

The stakes are high, and the clock is ticking. Will we be able to develop policies that address the existential risks associated with superintelligence before it’s too late? Or will we continue to rely on flawed economic models that ignore the most extreme and potentially catastrophic outcomes? The answer to these questions will determine not only our collective future but also the very fabric of human society itself.