Microsoft and Openai’s AGI battle is bigger than the contract

I learned first Regarding provisions from Microsoft CEO Satya Nadella. meanwhile An interview with him In May 2023, I asked about the agreement between Microsoft and Openai. I knew the contract had set a cap on how much profit Microsoft could benefit from the arrangement, so I asked if it reached that point and when and what would happen. The answer was a bit inexplicable.

“Essentially, their long-term thinking is that we’re going to be tight,” he told me. “If that happens, I think all bets are off, right?” He was almost unsure about the possibility, so I led him to wonder how serious he had taken it. “If this is humanity’s last invention, all bets are off,” he continued. “Different people will make different judgments about what it is and when it is.”

Until a few weeks later, I didn’t realize how important that resolve was. I’m working on it Openai featuresThe contract essentially declared that Microsoft cannot access the new model if Openai’s model achieves artificial general information. Otherwise, the terms of the contract, which had been extended until 2030, will be void. I wrote about it in my story, and although the clauses were not actually a national secret, it didn’t generate much debate.

That’s not the case anymore. This clause is at the heart of an increasingly frayed relationship between Microsoft and Openai and is currently being renegotiated. It was the subject of the research store in by information, Wall Street Journal, Financial Timesand yes, Wired.

However, the importance of the clause exceeds the fate of both companies that agreed to it. The tenuous terms of that contract lie at the heart of a fierce debate about changing the world and favours. If you want to understand what’s going on with AI, almost everything can be explained in terms of clauses.

Let’s dig into the details. Although the exact language is not publicly available, sources with knowledge of the contract confirm that there are three parts in the clause, each with its own meaning.

There are two conditions that Openai must be met to reject its technology to Microsoft. First, the Openai Committee determines that the new model has achieved AGI. This is defined on Openai’s charter as “a very autonomous system that surpasses humans in the most economically valuable work.” Fuzzy enough for you? It’s no wonder Microsoft is worried that Openai will make its resolve premature. The only way to oppose the Openai Commission’s declaration is to sue.

But that’s not all. You also need an OpenAI board to determine whether the new model has achieved “sufficient AGI”. This is defined as a model that can generate enough profits to reward other investors in Microsoft and Openai. Openai doesn’t really need to make those profits. Just provide evidence of that new model Intention Generate that prize. Unlike the initial decision, Microsoft must agree that Openai meets its standards but cannot unfairly challenge it. (Again, in the case of a dispute, the court can ultimately decide.)

Altman himself admitted to me in 2023 that the standards were vague. “It gives our board a lot of control to decide what will happen when we get there,” he said. In any case, if Openai determines that it has reached sufficient AGI, then there is no need to share these models with Microsoft. You don’t even need to use Microsoft’s cloud servers. Currently, Microsoft has the right to initially reject the work.