The grudge entanglement between Musk and OpenAI has added another touch.
On the evening of February 29, local time, Elon Musk (Elon Musk) sued OpenAI, CEO Sam Altman, and President Greg Brockman in court, and the indictment has been submitted to the San Francisco High Court.
Breach of contract, breach of fiduciary duty, unfair business competition, ......Musk made a total of five accusations, demanding that OpenAI open source, restrain profit-making, and conduct financial accounting. Musk has repeatedly criticized OpenAI for "deviating from its mission" and "forgetting its original intention", and now it is really written in black and white.
Mr. Musk co-founded and funded OpenAI with Mr. Altman and Mr. Brockman in exchange for relying on a founding agreement to ensure that artificial intelligence general intelligence (AGI) will benefit humanity rather than a for-profit company. As events unfold in 2023, his contributions to OpenAI are distorted to benefit the defendant and the largest company in the world. This is a serious betrayal of the founding protocol, subverting the original intent of the protocol and distorting the mission of OpenAI. ”
Although the majority of netizens already have an understanding of Musk's temper and style, some people still sincerely ask: Ah? Can this also be sued? The master of the meme has already arrived at the scene and presented the meme and the stalk**. In a meme, it was supposed to be openai, but it became "closedai", summarizing the news in one sentence.
Some people also dug out the original text of the complaint and found Huadian: Do you kid want to use the lawsuit to figure out what OpenAI's Q* is? Previously, it was rumored that OpenAI's new model, codenamed Q*, is the password to AGI (Artificial General Intelligence).
Meta and Google have released open-source models, and the "open" in OpenAI's name has become more and more ironic. In addition to being technically not open enough, in January this year, OpenAI was also found to have quietly changed its policy and no longer opened a large number of internal documents such as financial statements.
smashed open "openai, Musk punched first.
A read Musk's complaint, and the crocodile was going to shed tears.
The 35-page complaint tells a sad and aggrieved story: Musk, who is concerned about the survival of mankind, meets Ultraman in sheep's clothing. Tens of millions of dollars, gone; The danger of human extinction is even more serious.
In the complaint, there is a special section on "Mr. Musk's key role in OpenAI's start", saying that without Musk's efforts and resource investment, OpenAI is likely to never start. Not only did he donate $44 million between 2016 and September 2020 to rent and pay rent for OpenAI's initial office in San Francisco, but he also relied on his influence and connections to actively recruit talent. Ilya Sutskever, the later chief scientist of OpenAI, was snatched from Google by Musk himself.
But Musk's efforts are all based on OpenAI's founding agreement. This founding agreement is also the most critical fulcrum of Musk's lawsuit against OpenAI, which has two main points: non-profit nature and openness.
However, things started to change in 2017. In that year, Greg Brockman and others began to move OpenAI into a for-profit company. When Musk heard this, like this, the previous investment of real money ** turned out to be giving money for a start-up like a fool? In 2018, Musk left OpenAI.
By 2019, OpenAI established a profitable subsidiary. In 2020, OpenAI announced a cooperation with Microsoft, but the authorization is limited to the technology before AGI, that is, once AGI is implemented, it will no longer be licensed, and as for when AGI will be reached, it is up to OpenAI's board of directors to decide.
In a short period of a few years, although the two sides parted ways, they were still peaceful, until the emergence of ChatGPT and the development of GPT-4.
In the complaint, Musk sorted out the process of OpenAI's gradual closure: in 2018, OpenAI released the GPT model, announced the source**, and released the details**. In 2018 and 2019, OpenAI released GPT2 and GPT3 respectively, with accompanying research**.
Elon Musk
In March 2023, OpenAI released a new generation model, GPT-4, but the entire research and development process was kept secret. More critically, Musk believes that GPT-4 should be treated as an AGI model. The complaint cites evidence from an article published by Microsoft researchers stating that "given the breadth and depth of GPT-4's capabilities, we believe it could reasonably be considered an early (but still incomplete) version of an artificial general intelligence (AGI) system."
In other words, OpenAI did not keep its proprietary license to Microsoft on GPT-4 public in accordance with the founding agreement, and did not immediately terminate its exclusive license to Microsoft because it reached the AGI threshold. It's still the old saying of Musk - OpenAI has deviated from its mission.
An interesting point is that on the matter of "reaching AGI", Musk also made an accusation against Ultraman: after last year's OpenAI "coup", Ultraman, who was ousted by the former board of directors, returned to the company with the help of Microsoft and replaced almost all the board members. The complaint alleges that the new board members as a whole lack professional technical backgrounds and are unable to make a judgment on whether AGI has been met. In this way, OpenAI can postpone the announcement of AGI indefinitely, so as to continue to cooperate with Microsoft and safeguard business interests.
In short, this is a conspiracy matryoshka. Altman started with a non-profit organization, and it was a conspiracy to get Musk to pay for it, but in fact to turn to a profit-driven company. It is still a conspiracy to cooperate with Microsoft on the condition of "ex-AGI", but to postpone the announcement of the arrival of AGI in order to make the conditions meaningless.
bMusk sued OpenAI, what result do you want from the "Blue Sky Master"?
In the complaint, Musk filed five charges against OpenAI, including breach of contract, breach of trust business, and unfair business practices.
Among a series of demands, the core is the issue of money and open source.
Musk wants the court to rule that OpenAI will open source its technology and research results to the public, and ask the court to issue an injunction prohibiting OpenAI, Brockman and Ultraman, and Microsoft from profiting from the company's AGI technology.
In addition, Musk also asked for financial accounting to find out the amount of funds injected by Musk and other investors into OpenAI, and how much of these injections brought to OpenAI, and how much of it puffed up the pockets of the shareholders behind OpenAI and Ultraman.
Sam Altman
The complaint also states that injustice can only be avoided by enforcing the defendant's repeated promises. If there is no specific enforcement, then the defendant must at least pay compensation equal to the plaintiff's contribution.
In Musk's view, OpenAI started as a non-profit, took the donated money, and turned to commercial profit-driven, which would provide a toxic example for Silicon Valley.
Musk gave two examples to illustrate the seriousness of the matter. First of all, from a competitive point of view, it's like being able to score double points in a race. The complaint argues that for every dollar that an investor "invests" in a nonprofit organization through donations, he receives an income tax deduction of about 50 cents from the federal** and state** of the United States, so the cost is actually only 50 cents, while start-ups that start as traditional for-profit companies still pay a dollar.
Second, from the donors of "Generating Electricity for Dreams," "imagine donating to a nonprofit organization that claims to protect the Amazon rainforest, but then the organization forms a for-profit Amazon timber company that uses the donated funds to cut down the rainforest." ”
If the court endorses OpenAI's actions, then any startup looking to remain competitive in Silicon Valley will essentially have to follow this OpenAI playbook. This will become standard practice for start-ups, hurting legitimate nonprofits, taxes, and ultimately the people of California and beyond. It is worth noting that OpenAI's for-profit entity was recently valued at nearly $80 billion. ”
cIt seems that the whole world has noticed OpenAI's "non-open".
Since last year, the "open source legion" has been menacing. In March last year, Meta's Llama model was leaked, and open source community innovations emerged. Soon after, an article titled "We don't have a moat, and neither does OpenAI" went viral, written by researchers within Google. In July, Meta simply released LLAMA 2, which is open-source, free, and commercially available, with Microsoft as the preferred partner.
Open source becomes justice to some extent. Recently, Microsoft and the "new love" Mistral AI reached a multi-year cooperation, which has not only opened its own source, but also launched models directly on the seeds that people use to pirate content. Currently, Mstral AI has raised more than $400 million in funding at a valuation of about $2 billion.
In February this year, when OpenAI amazed the world with its Wensheng ** product Sora, Google also silently released the open-source model series GEMMA, providing a full set of model weights for free and explicitly allowing commercial use.
Open source is only open in the narrow sense.
In January of this year, Wired magazine found that OpenAI had quietly downgraded its "transparency". Since its inception, OpenAI has pledged to make its internal documents, including financial statements, available to the public. When Wired asked OpenAI for documents, it was denied: "We have financial statements on request. Since 2022, OpenAI's practices have been consistent with the standards and no longer share additional internal documents in the air. ”
Nowadays, OpenAI, which is not open source, surrounded by open source models, and whose transparency is decreasing, has become the target of netizens' memes due to the contrast between its performance and the "open" in its name. In many memes, OpenAI is spoofed as "ClosedAI", that is, from "Open AI" to "Closed AI".
Not counting the related memes, someone even made the official webpage of "ClosedAI", using the visual design of OpenAI's official website, and also issued a statement in the tone of Ultraman, with a full Yin and Yang Qi value: "By keeping our research closed, we will ensure that it is used for human welfare." ”
At least, compared to when ChatGPT just exploded and Musk scolded OpenAI for "deviating from its mission", there are more people on Musk's side now. Whether you can smash OpenAI with one punch depends on Mr. Musk's creation.