What is AGI (artificial general intelligence)? A straight answer might seem like trying to find a needle in a haystack, especially since the term has seemingly been turned into a buzzword with a different meaning each time you hear it. However, it is commonly defined as a sophisticated AI system that has surpassed human intelligence.
While it might seem like any other jargon, it’s actually an important part of Microsoft’s multi-billion-dollar partnership with OpenAI. A stringent clause in the agreement categorically highlights that the partnership will be severed once the ChatGPT maker achieves this coveted benchmark.
With the different variations in the meaning, it’s almost impossible to tell what constitutes AGI for Microsoft and OpenAI. However, a leaked document from earlier this year revealed that the term refers to an AI system that can generate up to $100 billion in profit.
It’ll be interesting to see if OpenAI can develop a system that can generate that kind of money amid bankruptcy claims and mounting pressure from investors to evolve into a for-profit venture to avoid outsider interference and hostile takeovers.
Microsoft’s partnership with OpenAI is seemingly taking a turn for the worse. Both parties are in high-stakes negotiations over the for-profit evolution as Microsoft attempts to protect its best interest and $14 billion investment in the AI firm.
However, the software giant is concerned that OpenAI might prematurely declare AGI. While Microsoft could seek legal action against the ChatGPT maker, citing bad faith, it could be trapped in the corridors of justice for an excruciatingly long period.
As highlighted by The Wall Street Journal, some Microsoft executives were skeptical about the stringent AGI clause in the agreement, but they agreed anyway because the company was lagging behind in the AI race.
Fast forward to now, OpenAI executives are reportedly in discussions to declare AGI by shipping an AI coding agent that supersedes the capabilities of an advanced human programmer (via The Information).
Earlier this year, Microsoft’s CEO Satya Nadella dismissed AGI milestones as “nonsensical benchmark hacking.” Interestingly, Sam Altman recently admitted that today’s computers and systems weren’t built for an AI world, potentially pushing the coveted AGI benchmark further away.
Microsoft and OpenAI’s partnership might have reached an impasse
Microsoft’s partnership with OpenAI is seemingly in the crosshairs, with the latter being on the verge of moving to court over anticompetitive behavior, as it tries to loosen Microsoft’s leash on its Intellectual Property, specifically a potentially $3 billion Windsurf acquisition. A separate report also reveals that Microsoft is asking for a lion’s share of OpenAI’s Public Benefit Corporation (PBC), which is way above what it is willing to offer.
Over the past few months, multiple reports have suggested that the once-best “tech bromance” in history is seemingly fraying. Microsoft has already started developing its own in-house models, potentially emancipating itself from an over-dependence on OpenAI for its technology.
Microsoft AI CEO Mustafa Suleyman indicated that the off-frontier models will be 3 to 6 months behind OpenAI’s, further indicating that the tech giant’s plan is to play a close second. In the same breath, Microsoft is also testing third-party models in Copilot.
If the ChatGPT maker moves forward with the plan to declare AGI, it could be well-positioned to sever its ties with Microsoft, cutting its IP rights and access to its flagship AI models.
Aside from funding, Microsoft provides OpenAI with cloud computing for its AI advances, though this arrangement was altered following the announcement of the $500 Stargate project designed to facilitate the construction of data centers across the United States.
Consequently, Microsoft lost its exclusive cloud provider status with Salesforce CEO Marc Benioff predicting that the tech giant won’t use OpenAI in the future. As a result, Microsoft has recently been seen bailing out of two mega data center deals because it didn’t want to provide additional support for ChatGPT training.
However, OpenAI CEO Sam Altman claimed that the company is no longer compute-constrained. Both companies have raised complaints about each other, with Microsoft indicating that GPT-4 is too slow and expensive to meet its customers’ needs, while OpenAI threw a fit about Microsoft not meeting its cloud computing needs, loosening its grip on AGI to other rivals.