The Information reports that Microsoft and OpenAI have a very particular definition of artificial general intelligence (AGI), which is profit driven. Where traditional technical and philosophical criteria for AGI fall short, the companies agree it is a milestone when the OpenAI systems generate at least $100 billion in profits. That said, on this metric, OpenAI is a long way from achieving AGI.
The agreement reportedly reached with them last year has big ramifications for their partnership. But currently, Microsoft enjoys privileged access to OpenAI’s latest cutting edge AI models. But then, Microsoft loses this, once OpenAI achieves AGI under this profit-centric definition. That puts you in a peculiar dynamic, with the financial milestone far more than the technological one serving as the barometer for your partnership’s future
However, OpenAI is very much facing funds. But it’s likely to lose billions of dollars this year, and not expect to make any money until 2029. The $100 billion profit threshold is retreating further and further from the realities of these financial struggles. It’s another decade, or more, that it will take Microsoft to realistically access OpenAI’s technologies in any useful form, despite claims that actions by OpenAI might prematurely declare AGI, in hopes of leaving Microsoft in the dust.
The discussion around recent o3 model of OpenAI’s only serves to highlight the complexities of this financial definition of an AGI. AI models that outperform other AI models are o3 and while it is allegedly better than other AI models, their high computational costs bring the question of profitability into play. That is why Microsoft and Openai have found agreement at the level of digital profits; something that may be hard to synchronize with the organic pace of technological development.
At the same time, it presents important questions about how technological progress is envisioned as a business. The $100 billion AGI benchmark, although not conventional by any means, does suffice to give the AGI a concrete number it can be marked down to, while standing at odds with the usual definition of AGI as human like reasoning, or plasticity. The Microsoft OpenAI partnership sits in a precarious standing, given the vagueness of the financial and technical paths to AGI for now.