A lawsuit brought against one of the fastest growing companies in history by the world's richest man is inevitably interesting. However, while the allegations have not yet been proven, the incident has already exposed a series of emails between Elon Musk, Sam Altman and others in the early days of OpenAI. Here are some of the more interesting snippets I found while sifting through their correspondence.
Note that these emails were published as part of an attempt to prove that OpenAI is somehow violating antitrust laws (a frankly implausible claim). Musk has also made some sense of his feelings of betrayal when OpenAI abandoned its original vision of being a nonprofit organization led by Tesla's CEO.
They don't tell the whole story, but they're still interesting in their own right.
Perhaps the most interesting email was the one in which former chief scientist Ilya Satskeva explained the team's anxiety to install Musk as the company's leader.
The current structure provides a pathway to unilateral and absolute control over AGI. [artificial general intelligence]. You have stated that you do not want to control the final AGI, but during this negotiation you have indicated to us that absolute control is very important to you.
As an example, you said you need to be the CEO of a new company so everyone knows you're in charge, but you don't like being the CEO and would rather be the CEO. I also said that there is no.
Therefore, we are concerned that as your company makes real progress toward AGI, you may choose to retain absolute control of your company despite current intentions to the contrary. Masu.
OpenAI's goal is to make the future better and avoid AGI dictatorship. you are worried about demis [Hassabis, at Google-owned DeepMind] There is a possibility of creating an AGI dictatorship. So do we. So it's a bad idea to create a structure where you can become a dictator if you want to. Especially considering that you can make another structure that avoids this possibility.
This is not entirely about corporate management. Sutskever is concerned about the existential threat of AI being created by just one human.
Mr. Sutskever also expressed concern about Mr. Altman, using language similar to the one the board would later use while accusing him of not being “consistently candid.”
Because we don't understand your cost function, we couldn't fully trust your judgment throughout this process.
We don't understand why the title CEO means so much to you. The reason you stated has changed, but it's hard to really understand what caused it.
Is AGI really your main motivation? How does it relate to your political goals?
Given past developments and Altman steering the company toward a more traditional enterprise SaaS position, it seems like his goals were more business than philosophy.
One interesting tidbit is that as of 2017, OpenAI was seriously considering acquiring or somehow merging with chipmaker Cerebras to use Tesla's resources in some way. Mr. Sutskever said:
I feel strongly that if we decide to acquire Cerebras, it will be done through Tesla.
In the end, they didn't do it, but the email doesn't say why.
Incidentally, this was at a time when Musk was looking to make OpenAI one of his many assets, and leaders were open to the possibility. OpenAI co-founder Andrej Karpathy writes:
The most promising option in my opinion is, as I mentioned earlier, for OpenAI to belong to Tesla as a cash cow. […] If we do this really well, the transportation industry is large enough that we can drive Tesla's market cap to high levels of O(~100,000) and use the proceeds to fund AI development at an appropriate scale.
Again, this didn't happen for a number of reasons, as is obvious in hindsight. Indeed, while Tesla's market capitalization has increased, the self-driving aspect (which Mr. Karpathy later tried to accelerate when he took a job at Tesla) has proven more difficult than expected, and Tesla's earnings still remain has not made a meaningful contribution.
In terms of making money, Microsoft has been on board since early 2016, providing $60 million worth of compute on Azure to OpenAI in exchange for both companies to “promote” each other. No one is interested in this kind of corporate backbiting, which Musk wrote is “sickening.”
They ended up paying a much higher amount, but there was no obligation on either side. “It's going to be worth more than $50 million, just so you don't look like a Microsoft marketing bastard,” Musk wrote.
Finally, an important point mentioned by board member Siobhan Gillis (who later became the mother of Musk's three children): Valve founder Gabe Newell was an early donor to the project. In addition, Altman and Greg Brockman's “informal advisory board.” It is unclear what role he had on a day-to-day basis there. Mr. Newell has been contacted for comment.