I learned first Terms about Microsoft CEO Satya Nadella. In an interview with him in May 2023, I asked about the deal between Microsoft and Openai that granted his company exclusive access to the startup’s groundbreaking AI technology. I know the contract has set limits on how much profit Microsoft makes from the arrangements, and I asked him what would happen if he reached this and when he did. The answer is a bit confusing.
“Their long-term idea is fundamentally that we’re going to do super smart,” he told me. “If that happens, I think all the bets are closed, right?” He seemed almost harsh about the possibility, which made me wonder how much he values it. “If this is the last invention of mankind, then all bets are closed.” “Different people will have different judgments about what that is and when.”
It wasn’t until a few weeks later that I realized how important that determination would be. I understand about the features of OpenAI, and I understand that the contract basically announced that Microsoft will no longer have access to its new model if Openai’s model gets artificial general intelligence. Otherwise, the contract terms that will be extended to 2030 will be invalid. Although I wrote about it in the story, and the clause was never a secret to a country, it didn’t cause much discussion.
This is no longer the case. The clause is at the center of an increasingly worn relationship between Microsoft and Openai and is being renegotiated. Through information, the Wall Street Journal, the Financial Times and Wired, it has been the subject of investigating stores.
But the meaning of this clause goes beyond the fate of the two companies that agree to the clause. The tough conditions of the contract are at the heart of a fierce debate about achieving the change (and lucrative) of the world, if the world is realized, and what control over profit-driven companies a technology that makes Sauron’s circle of power look like a plastic Doodad in a store of the business. If you want to know what’s going on in AI, almost everything can be explained by terms.
Let’s study the details. Although the exact language has not been disclosed, sources who understand the contract confirm that the clause has three parts, each with its own influence.
Openai must meet two conditions to deny its technology rejection from Microsoft. First, the OpenAI board will determine that its new model has implemented AGI, which is defined in Openai’s charter as “a highly autonomous system that surpasses humans in most economically valuable work.” Is it vague enough for you? No wonder Microsoft is worried that Openai will make a decision too early. The only way to oppose the Openai board statement is to file a lawsuit.
But that’s not all. The OpenAI board of directors is also needed to determine whether the new model has received “sufficient AGI”. This is defined as being able to generate a model that is sufficient to reward other investors in Microsoft and Openai (more than $100 billion). Openai does not have to actually make these profits, only provides evidence to its new model Will be Generate bounty. Unlike the first determination, Microsoft must agree that Openai meets the standard, but cannot object unreasonably. (Again, if the court may decide eventually.)
Ultraman himself admitted to me in 2023 that the standards are vague. “This gives our board a lot of control to decide what happens when you get there,” he said. In any case, if Openai thinks it has reached enough AGI, it doesn’t have to share these models with Microsoft, which Microsoft will be bothered with earlier versions that are now outdated. It doesn’t even have to use Microsoft’s cloud servers; Microsoft currently has the right to reject the work in the first place.
