Elon Musk sues OpenAI, Sam Altman, and Gregory Brockman: OpenAI’s GPT-4 has breached the threshold of reasonable AGI, while the company has become a tool to make Microsoft more money, totally going back on its original aim to fight Google in the race to AGI and be open for all humanity’s benefit.
Older AI used to be good in specific tasks, such as IBM’s Deep Blue which defeated Kasparov in chess1. It could only play chess. Google, in 2014, acquired the deep learning research group DeepMind and soon released AlphaZero, another program really good at beating chess grandmasters. But AlphaZero was different. It used reinforced learning (incentive on reward). Given the rules (and no other knowledge) of other games like shogi and Go, AlphaZero could also topple world champions in those games2. This was alarming and Elon Musk was one of the people who believed AGI or artificial general intelligence is closer than ever before, posing a very real threat. In the hands of a closed, for-profit company like Google that already has vast amounts of data, this could be bad for humanity at large.
So, in 2015, Sam Altman, another Silicon Valley entrepreneur worried about the risks of AGI, got in touch with Musk. Altman’s proposal was simple – make our own AI lab to “catch up to Google in the race for AGI” but by being completely opposite, or open, for the greater good. The new, non-profit lab to develop AGI for humanity’s good would be called OpenAI3.
A series of events later, Elon Musk steps down from OpenAI4, the company is instead backed by another closed, for-profit tech giant (Microsoft)5, and has pretty much discarded its founding principles, such as “The corporation is not organized for the private gain of any person.”
OpenAI is not open anymore.
Musk voiced his criticism of this numerous times. The singular event that caused the biggest ripples in the pond was Microsoft’s large cash injection into the ex-open AI lab. With that came a push toward launching tools and products to acquire customers and charge them. ChatGPT, of course, was that hero product, ultimately catapulting the whole tech industry on the race course.
It’s all come back to haunt them in the latest lawsuit filed individually by Elon Musk against OpenAI (more specifically, two of the founding members and 8 different companies that make up the OpenAI corporate structure). The lawsuit outlines how Musk was a major moving force behind the then-open OpenAI’s decision-making and contributed a majority of its funding in the first several years, advised on research directions, and recruited the top talent. With Musk’s money, the lawsuit comments, it wouldn’t have been possible to get the top talent for AI because Google was offering much better money.
The lawsuit also makes a point that GPT-4 is AGI and that Microsoft’s own researchers concur that it’s an early and incomplete version of an AGI system, better in reasoning than the average human. It also talks about Q*, an internal AGI project by OpenAI.
Apparently, Microsoft only has a license to OpenAI’s pre-AGI or non-AGI models and tools (didn’t know about that nifty distinction). But since there is no clear metric to tell if it’s AGI or not, it’s completely up to the board’s discretion to decide whether a model has reached AGI or not.
The point is that the company (which also went through a coup) has already achieved AGI and Microsoft is making a profit from that. Thus, the original agreement on which the company was founded went up in flames. Instead of being for the benefit of all humanity, the lawsuit alleges, the company is a “closed-source de facto subsidiary of the largest technology company in the world: Microsoft” that’s just trying to maximize the profits for Microsoft.
The lawsuit’s objective is to compel OpenAI to adhere to its original agreement and get back to its mission to develop AGI for humanity’s benefit, not Microsoft’s.
The Hero?
In 2013, Musk tried buying DeepMind over Google in fears of AGI that could harm humanity, but failed. He talked with Obama about the risks of AI now that Google had DeepMind, but regulations never came. He then helped start OpenAI to champion AGI-for-humanity and fight against DeepMind, but that failed as well. In 2023, he called for a pause on AI development with the support of 1000+ tech leaders6, but that seems to have created nothing apart from hollow organizations.
For Musk, it’s now time to take them to court and force their hand legally. His lawsuit foreshadows that humanity is threatened. To some, it might be a clear story of a hero, a champion of humanity fighting against tech giants (while being one himself).
To others, it’s yet another ploy to get more attention after the $16/month Grok has more or less fizzled out of relevance7.
Or the Villain?
Musk stepped down from the OpenAI board of his own volition in 2018 due to a conflict of interest because Tesla was developing its own AI4 – so he’d be competing for talent. Apparently, Musk’s love for humanity’s protection was not bigger than a conflict of interest over self-driving car AI. He also poached talent off OpenAI8.
Being a non-profit entity, OpenAI relied on investments. It was announced that even though Musk is leaving OpenAI, he has promised to keep the funding coming, something to the tune of $1 billion in total. But the funding just stopped after he left. There was basically no money to work on AI anymore. And that’s why OpenAI at first created a for-profit entity so they could raise some money. A tech company is always willing to invest in a tech startup in return for profits or equity. Only a for-profit company is worth anything by this logic.
And this got the ball rolling – Microsoft noticed this company and poured its cold, hard cash. A whopping $13 million of it, between 2019 and 2023 without taking any stake, just a share in the profits.
The dams were opened. The flood soon came.
To pay Microsoft’s profits back, OpenAI now has to make money. It can’t make money unless it sells a product to consumers. Now, Musk doesn’t like this. So, he’s telling the folks at OpenAI to get back to being a non-profit without explaining how this $13 million will be paid back with just the non-AGI GPT-3, which is yesterday’s news. And how can we be truly confident that GPT-4 is AGI? Whose definition is correct in an industry without a product?
By the way, remember why Musk had to leave OpenAI, forsaking humanity? As it turns out, in 2024, Tesla’s autonomous driving AI is still at Level 2. There are 6 levels of autonomy in self-driving cars, and companies like Audi have already surpassed Tesla by reaching Level 3, and several companies are already working on Level 4 prototypes9.
1. “In 1997, an IBM computer beat a chess world champion for the first time” – CNN Vault (link)
2. “Google DeepMind’s Go AI Has Mastered Chess And Shogi” – Sam Shead, Forbes (link)
3. “Elon Musk + Sam Altman Launch OpenAI Nonprofit That Will Use AI To ‘Benefit Humanity’” – Daniel Terdiman, Fast Company (link)
4. “Elon Musk, who has sounded the alarm on AI, leaves the organization he co-founded to make it safer” – Jordan Novet & Lora Kolodny, CNBC (link)
5. “Microsoft’s $10 Billion Bet on OpenAI Is Now Up in the Air” – Anthony Di Pizio, The Motley Fool (link)
6. “Elon Musk and Others Call for Pause on A.I., Citing ‘Profound Risks to Society’” – Cade Metz & Gregory Schmidt, The New York Times (link)
7. Elon Musk’s AI company xAI released the Grok-1 model that’s apparently the best in its class (GPT-3.5) and inferior to models trained with more data, like Palm 2, Claude 2, and GPT-4 as per the chart on xAI’s official page. However, after 4 months of its launch, it’s still a part of no research paper focused on AI chatbots or language models. What’s more, it might not actually be funny and could’ve been a rip off of ChatGPT, to begin with (link, link), sometimes regurgitating ChatGPT responses verbatim, causing some to think that it’s possibly hooked to a ChatGPT API.
8. “Elon Musk has poached a top mind in AI research—from himself” – Dave Gershgorn, Quartz (link)
9. “The 6 Levels of Vehicle Autonomy Explained” – Synopsis (link)
