Tuesday, March 5, 2024

Is ChatGPT-4 Artificial General Intelligence?

Hitesh-choudhary-t1PaIbMTJIM-unsplashYesterday I wrote about some problems with the alleged Founding Agreement that makes up the foundation of Elon Musk's lawsuit against OpenAI and Sam Altman. Today I want to shift to one of the suit's prayers for relief.

Musk wants the court to declare that ChatGPT (and that Q* and/or OpenAI's next generation large language models) constitute "artificial general intelligence" and are thus "outside the scope of OpenAI’s license to Microsoft."

This prayer for relief requires a little unpacking. Earlier in the complaint, Musk says that OpenAI's license agreement excludes artificial general intelligence. That is, if ChatGPT-4 is artificial general intelligence, OpenAI cannot license it to Microsoft. And, Musk says, Microsoft's own researchers believe that ChatGPT-4 "could reasonably be viewed as an early (yet still incomplete) version of an artificial general intelligence . . . system."

That seems bad, on the surface. Maybe the parenthetical helps OpenAI? Is an incomplete artificial general intelligence system different from an excluded artificial general intelligence system?

It turns out, though, that it doesn't matter. Musk doesn't claim that "artificial general intelligence" is a defined term in the alleged Founding Agreement. Rather, in his complaint he quotes OpenAI's website as saying that " OpenAI’s Board 'determines when we’ve attained AGI.'”

And, um, that's not great for his assertion. See,OpenAI's board seems to have decided that ChatGPT is not artificial general intelligence, and can thus be licensed to Microsoft.

Clearaly Musk disagrees with the board. And if the judge disagrees?

It doesn't really matter. Under Delaware law, a "court cannot second-guess the wisdom of facially valid decisions made by charitable fiduciaries, any more than it can question the business judgment of the directors of a for-profit corporation." That is, the board's decision is protected by the infamous (and powerful) business judgment rule. In essence, as long as the board appropriately informed itself, a court won't second-guess its decision. If the board, which Musk acknowledges has the authority to determine whether an AI technology represents artificial general intelligence, says ChatGPT-4 is not, for legal purposes, it's not. Even if Microsoft employees believe it is. Even if Elon Musk believes it is.

Samuel D. Brunson

Photo by Hitesh Choudhary on Unsplash

https://lawprofessors.typepad.com/nonprofit/2024/03/is-chatgpt-4-artificial-general-intelligence.html

Current Affairs | Permalink

Comments

Post a comment