Microsoft and OpenAI Define AGI: Financial Details Revealed

OpenAI's AGI Definition Tied to Profitability

A recent report from The Information reveals that Microsoft and OpenAI have established a precise, internal benchmark for artificial general intelligence (AGI). This definition is surprisingly centered around the startup’s financial performance.

According to the report, an agreement signed between the two companies last year stipulates that OpenAI will only be considered to have achieved AGI when its AI systems generate a minimum of $100 billion in profits annually.

A Departure from Traditional AGI Definitions

This profit-based criterion significantly diverges from the more conventional, technically focused and philosophical understandings of AGI. Many experts anticipate a more rigorous assessment of true general intelligence.

Currently, OpenAI is projected to incur substantial losses this year. The company has communicated to investors that profitability is not anticipated until 2029.

Implications for Microsoft's Access to OpenAI Technology

The timing of AGI achievement is crucial because Microsoft’s access to OpenAI’s technology is contingent upon the startup reaching AGI. The term AGI remains somewhat ambiguous, leading to varied interpretations.

Speculation has arisen that OpenAI might prematurely declare AGI to limit Microsoft’s future control. However, this agreement suggests that Microsoft could retain access to OpenAI’s models for potentially another decade or longer.

The Cost of Advanced AI Models

Recent discussions surrounding OpenAI’s o3 model and its potential as a step towards AGI have been ongoing. While o3 demonstrates improved performance compared to other AI models, it also requires considerable computational resources.

These high compute costs present a challenge to OpenAI and Microsoft’s profit-driven definition of AGI. Successfully reaching the $100 billion profit target will require significant efficiency gains.