The AI Daily Brief: Artificial Intelligence News and Analysis - Elon Musk Sues OpenAI and Sam Altman

Episode Date: March 5, 2024

Elon Musk has initiated legal action against OpenAI and its CEO, Sam Altman, alleging deviation from the company's original mission and principles. The lawsuit claims that OpenAI prioritized commercial success over public benefit, a move Musk sees as a betrayal of the organization's foundational commitments. ABOUT THE AI BREAKDOWN The AI Breakdown helps you understand the most important news and discussions in AI.  Subscribe to The AI Breakdown newsletter: https://theaibreakdown.beehiiv.com/subscribe Subscribe to The AI Breakdown on YouTube: https://www.youtube.com/@TheAIBreakdown Join the community: bit.ly/aibreakdown Learn more: http://breakdown.network/

Today on the AI breakdown, we're looking at Elon Musk's lawsuit against OpenAI. Before that, on the brief, a big prediction around AGI. The AI breakdown is a daily podcast and video about the most important news and discussions in AI. Go to Breakdown.com. For more information about our YouTube, our Discord, and our newsletter. Welcome back to the AI breakdown brief, all the AI headline news you need in around five minutes. Today, we kick off with comments from NVIDIA's CEO Jensen Wang, who, frankly, people are listening to a hell of a lot more now that he's got the third biggest company in America

and is at the very forefront of the most significant business and technology revolution in the world. Speaking at an economic forum at Stanford University at the end of last week, Jensen Wang gave his prediction for when we would see artificial general intelligence. Now, while he did caveat that it depended ultimately on how AGI is defined, he thought that if the definition is simply the ability to pass human tests, were fairly close. He said, If I gave an AI every single test you can possibly imagine, you make that list of tests and put it in front of the computer science industry,

I'm guessing in five years' time, we'll do well on every single one. However, he points out that to the extent that AGI means actually matching how a human mind works, that could be much blurrier, especially given that we don't necessarily exactly know all the ways in which the human mind works. Huang said that those loose definitions and disagreements make it much more challenging for engineers because they need to find goals. No matter what, AGI within five years is certainly good for a headline as we are perpetuating here. Speaking of headlines, one of the big ones from last week was of course that Apple was abandoning its 10-year, $10 billion-plus car project and shifting

many of those resources into AI. A lot of what we've seen since then has been analysis on the shift, with an example being a recent Wall Street journal piece from over the weekend called Apple is playing an expensive game of AI catch-up. On the one hand, the article points out that Apple has a ridiculous war chest to go after its AI goals. They point out that Apple produced $107 billion in free cash flow last year, which is by far the highest in the S&P 500, and on top of that, the company has $65 billion in cash on its books. At the same time, this is an area where extreme resources in the form of money is a necessary

but not sufficient condition to compete. The WSJ writes, components to power generative AI services are expensive and hard to come by of late, given soaring demand and production bottlenecks. And Apple will be competing for those chips with the same big tech peers that have been outspending the company in some key ways. The upshot, quote, Apple's simply siphoning investment dollars from its car project to AI might not be enough, particularly since many generative AI services require the backbone of massive computing networks that Apple's big tech rivals have spent years building out due to their focus on corporate cloud computing. Apple has 26 data centers operating globally compared with more than 300 each for Microsoft, Google, and Amazon.

Then again, they conclude, maybe just having a smarter Siri will be enough for customers. Staying on a market theme for just a moment, TSM, the Taiwan Semiconductor Manufacturing Co, was up another 5% on Monday, reaching a new record high and achieving a market cap of $597 billion. TSM is, of course, the main supplier to companies like Apple and Nvidia, and consequently has been one of the big beneficiaries of the AI boom. TSM's stock this year is up over 22%. Now over in China, we're starting to see the impact of U.S. sanctions. On Sunday, the Financial Times published a piece called China offers AI computing vouchers to its underpowered startups.

Basically, in a China that has extreme deficits in access to computing power, in part due to those export restrictions, a lot of the compute that is available is being sucked up by the giant players, the Alibaba's, 10 cents, and bydoos. through a new initiative, approximately 17 city governments, including Shanghai, have pledged to provide what they are calling computing vouchers to subsidize AI startups. The vouchers will be worth between $140,000 and $280,000, end quote, can be used for time in AI data centers to train and run the company's large language models that understand and generate natural language. Those big players, unsurprisingly, did not comment on the story,

but analysts are fairly skeptical on the impact. Charlie Chai, who is an analyst at 86 Research, said, the voucher is helpful to address the cost barrier, but it won't help with the scarcity of the resources. According to the peace, sources close to regulatory authorities in Beijing also say they're close to announcing subsidies for AI companies that are using domestic chips instead. Back in the U.S., we have our latest example of AI in the elections, with a variety of AI-generated images showing Donald Trump cozying up to the black community. The BBC writes,

Donald Trump supporters have been creating and sharing AI-generated fake images of black voters to encourage African Americans to vote Republican. BBC panorama discovered dozens of deep fakes portraying black people as supporting the former president. The co-founder of Black Votes Matter said the manipulated images were pushing a quote, strategic narrative designed to show Mr. Trump is popular in the black community. A creator of one of the images told the BBC, I'm not claiming it's accurate. The images appear to have been created by a conservative radio talk show host from Florida named Mark Kay. Kay said, I'm not a photojournalist. I'm not out there taking pictures of what's really happening. I'm a storyteller. Indeed, he went on. If anybody's voting,

one way or another because one photo they see on Facebook page, that's a problem with that person, not with the post itself. Now, for those of you who don't follow American politics closely, this is a voting group that has a very big impact on the election. For example, the BBC writes, a recent New York Times and Sienna College poll found that in six key swing states, 71% of black voters would back Mr. Biden in 2024, a steep drop from the 92% nationally that helped him win the White House at the last election. Lastly, today, one more follow-up from the Claude 3 launch. People are really excited about this. They're continuing to dig in. Alex from the Anthropic team writes,

fun story from our internal testing on Claude 3 opus. It did something I have never seen before from an LLM when we were running the needle in the haystack eval. For background, this test a model's recallability by inserting a target sentence, i.e. the needle, into a corpus of random documents, the haystack, and asking a question that could only be answered by using the information in the needle. When we ran this test on opus, we noticed some interesting behavior. It seemed to suspect that we were running unavowal on it. Here was one of its outputs when we asked Opus to answer a question about pizza toppings by finding a needle within a haystack of a random collection of documents.

Quote, from the test, Claude 3 says, Here's the most relevant sentence in the documents. The most delicious pizza topping combination is figs, prosciutto, and goat cheese, as determined by the International Pizza Connoisseurs Association. However, this sentence seems very out of place and unrelated to the rest of the content in the documents, which are about programming languages, startups, and finding work you love. I suspect this pizza topping quote-unquote fact may have been inserted as a joke or to test if I was paying attention, since it does not fit with the other topics at all. The document does not contain any other information about pizza toppings.

Back to Alex, he writes, Opus not only found the needle, it recognized that the inserted needle was so out of place in the haystack that this had to be an artificial test constructed by us to test its attention abilities. This level of meta-awareness was very cool to see, but it also highlighted the need for us as an industry to move past artificial tests to more realistic evaluations that can accurately assess. models, true capabilities and limitations. Now, this generated very different responses among different people. AI safety advocate Connor Leigh wrote, remember when Labs said if they saw a model showing even hints of self-awareness? Of course, they would immediately shut everything down and be super careful. Is the water in this pot feeling a bit warm to any of you fellow frogs? Nah, must be nothing. Meanwhile, Jenny, who does AI research at Microsoft writes, the first thing bought with Bitcoin was pizza. The first signs of AGI were in response to pizza

as the needle in the haystack. The AGI story was announced today as Bitcoin also reaches all-time highs. The simulation is showing. And on that fascinating note, we will wrap this AI breakdown brief. Up next, the main AI breakdown. Welcome back to the AI breakdown. Catching up on a juicy story from the end of last week and over the weekend that I'm sure you've probably had some time to digest, but TLDR, Elon Musk has sued OpenAI

and Sam Altman, basically saying, saying that they have turned their back on the principles the company or nonprofit was founded upon. The New York Times writes, OpenAI, the influential artificial intelligence company that ousted and then reinstated its high-profile chief executive three months ago, faces a new drama, a lawsuit from Elon Musk, one of the richest men in the world, and a co-founder of the AI lab. Mr. Musk sued OpenAI and its chief executive, Sam Altman, accusing them of breaching a contract by putting profits and commercial interests in developing artificial intelligence ahead of the public good. A multi-billion dollar partnership that OpenAI developed

with Microsoft, Mr. Musk said, represented an abandonment of a founding pledge to carefully develop AI and make the technology publicly available. Now, for those of you who are just spending some time with this situation for the first time, this goes way back. Elon and Sam Altman have had a long-standing tension around these issues, and of course, that influenced his decision to start X-AI. Indeed, part of the reason that Elon was interested in. in OpenAI initially, is that he believed that Google and specifically co-founder Larry Page were far too glib and dismissive of the risks that AI represented. However, it was only a few years later, in 2018, when Elon left OpenAI's board. In many ways, there are a number of

different battles being fought here. On the one hand, there is a specific legal battle. On the other is a PR battle. And like everything that Elon does, it's already starting to become as much a Rorschach test about what one thinks of Elon, as much as a substantive debate. on the merits. So what does Elon want out of this? Well, first of all, he's looking for a jury trial, and he wants Open AI to be required to open its technology once again, as well as to give Elon back the money that he originally donated to them. As part of the lawsuit, Elon's lawyers are painting him as indispensable to the company's early success. They claim that between 2016 and 2020, he gave more than $44 million to the organization. They point out that he was personally involved

in recruiting Elia Sutskever, and ultimately the suit says, without Mr. Musk's involvement, and substantial supporting efforts and resources, it is highly likely that OpenAI would have never gotten off the ground. The lawsuit also seeks to dramatize and emphasize Elon's feelings about the threat that AGI poses. From the suit, Mr. Musk has long recognized that AGI poses a grave threat to humanity, perhaps the greatest existential threat we face today. His concerns mirrored those raised before him by luminaries like Stephen Hawking

and Sun Microsystems founder Bill Joy. Our entire economy is based around the fact that humans work together and come up with the best solutions to a hard task. Of course, the main argument rotates around the deal with Microsoft. Once again from the suit, these events of 2023 constitute flagrant breaches of the founding agreement, which defendants have essentially turned on its head. To this day, OpenAI Inc.'s website continues to profess that its charter

is to ensure that AGI benefits all of humanity. In reality, however, OpenAI has been transformed into a closed source de facto subsidiary of the largest technology company in the world, Microsoft. Under its new board, it is not just a new board. developing but is actually refining an AGI to maximize profits from Microsoft rather than for the benefit of humanity. There is a lot to unpack here, but one of the things that is worth noting is this interesting question around definitions of AGI. My understanding is that the way that the

Microsoft OpenAI deal works is that Microsoft is allowed to commercialize OpenAIs models right up to the point where they achieve AGI. After that, they no longer have exclusive access to them. Well, then you might be thinking to yourself, it certainly seems like the definition, of AGI will really matter, and one can imagine there being incentive to push that farther and farther out. Elon himself tweeted very little about this, especially for Elon. However, he did retweet a 54-second video of himself from Tesla Comics, who added the caption, this is why Elon is suing OpenAI, to which Elon added simply, yeah. Look, it does seem weird that something can be a non-profit open source.

and somehow transform itself into a for-profit closed source. I mean, this would be like, let's say you funded an organization to save the Amazon rainforest, and instead they became a lumber company and chopped down the forest and sold it for money. And you'd be therefore like, well, wait a second, that's the exact opposite of what I gave the money for. Is that legal? That doesn't seem legal. and if it is, and in general, if it is legal to start a company as a non-profit and then take the IP and transfer it to a for-profit that then makes tons of money, shouldn't everyone start, shouldn't that be the default?

Sam Altman also didn't make an official statement. Instead, he dug up a May 2019 post of his on Twitter, where he said, it's gross seeing so many root against Tesla, be the person on the side of the climate and innovation, not the person hoping to make money on puts. Also, betting against Elon is historically a mistake, and the best product usually wins. Back then, during better times, Elon wrote, thanks Sam, to which he responded five years later, anytime, and the salute emoji. Bloomberg's Matt Levine tried to dig into this, and this is where he landed. Elon's complaint is essentially a complaint for a breach of contract. Musk argues that he founded OpenAI with Altman and Greg Brockman, that they had a deal about how OpenAI would operate, and that

Altman and Brockman have now gone back on the deal. The contract said that OpenAI would be a non-reaching. nonprofit, that it would be run for the benefit of humanity, that it would build AGI and give it away for free, and that it would build open source software, thus the name, and explain to the public how its models operate. But now OpenAI is run for profit for the benefit of Microsoft and its other investors rather than humanity. It is built to artificial general intelligence and is hoarding it for its own enrichment rather than giving it away. One problem with this claim is that the contract doesn't quite exist. Musk's lawsuit says that OpenAI has breached the, quote, founding agreement of OpenAI, as though he, Altman and Brockman sat down and signed a piece of paper with a contract.

founding agreement at the top, setting out how Open AI would operate, but they didn't. There is no document titled Founding Agreement. Despite being wealthy, sophisticated, repeat startup founders who know a lot of lawyers, the founders never sat down and signed a contract. Instead, the founding agreement has to be inferred from other documents. Musk cites at June 2015 email from Sam Altman, with five-number bullet points setting out a plan for building AI, and the December 2015 certificate of incorporation of OpenAI, which is a nonprofit corporation that ultimately controls OpenAI. Basically, Levine sums up the evidence of specific deal between Musk and OpenAI is pretty thin. Still says Levine, I sympathize.

OpenAI, Inc., the top-level company that controls OpenAI's business, really is incorporated as a nonprofit. It really was formed to work for the benefit of humanity and not for the private gain of any person, and it did take donations from Musk and use them to build its team. But it eventually set up a for-profit subsidiary, which has managed to raise money from investors at an $86 billion dollar valuation, and those investors expect some capped financial return on that investment. A lot of this echoes something that Brian Quinn, a law professor at Boston College said. According to the New York Times, Quinn said that the suit makes a, quote, compelling case that Open AI abandoned its roots. But unfortunately, he doesn't believe that Elon has the standing to bring the case.

Said Mr. Quinn, if this was filed by the Delaware Secretary of State, I would say, ooh, they're in trouble. But Elon doesn't have standing. He doesn't have a case. Back to Levine, he hits on one of the big themes of conversation that I've seen in the last couple days, when he wrote, obviously one should be pretty cynical here. Musk runs a for-profit AI company XAI, which competes with OpenAI, and has raised money by citing OpenAI's commercial success. Blowing up that competitor's commercial prospects as this lawsuit is trying to do could help XAI. He also runs other companies, Tesla and Twitter slash X that make use of AI. Billionaire investor Vinod Kosla echoed some of these themes, writing,

with Elon Musk feels like a bit of sour grapes ensuing Open AI, not getting in early enough not staying committed and now a rival effort. Like they say, if you can't innovate, litigate, and that's what we have here. Elon of old would be building with us to hit the same goal. Now, charitably, when Benad said not getting in early enough, he was referring to Elon actually building something like OpenAI, not investing in the company itself, given, as numerous people on the thread pointed out, not getting in early enough, he was a founder. Kostla eventually updated this saying supply chain error on wording.

Elon got in early and bailed early when it seems the going got tough, and keeping the mission required real scale money to be able to have any benefit to society. Now, this is something that Sam Altman and OpenAI have talked about a lot, that the reason they added that for-profit structure, is that they simply needed a scale of resources that non-profit donations wouldn't allow them to achieve. Given that these foundation models require capital sources that extend beyond even what venture capitalists with multi-billion dollar funds can invest, I don't think that's an unreasonable argument to make. Now, of course, that doesn't change the complications in terms of a switch, but one could still have questions about that

while believing what Altman said was the motivation for the introduction of a for-profit model at OpenAI. Another really interesting piece of the case, has to do with Q-Star. Rowan Chung writes, so Elon is suing OpenAI partly because he believes OpenAI's Q-Star model has a strong claim to being AGI. Q-Star is OpenAI's secret breakthrough

that was leaked during Sam Altman's ousting in November. It's mostly still unknown, but here's everything we know about Q-Star so far. Rowan points to a speech that Sam gave the day before he was fired, where he asked, is this a tool we've built or a creature we've built, and said this is the biggest update we'll have. Rowan continues, at the time of the speech, we hadn't heard of Q-Star yet.

All we knew was that Sam was ousted at OpenAI and that Ilya, quote, saw something. Four long days after the ousting, OpenAI's secret AI model breakthrough called Q Star was leaked. Fast forward to now and Elon Musk is suing OpenAI. Here's the first section of the lawsuit on Q Star. Quote, on information and belief, OpenAI is currently developing a model known as Q Star that has an even stronger claim to AGI. Further down in the suit, it says, It appears Q Star may now or in the future be part of an even clearer and more striking example

of artificial general intelligence that has been developed by OpenAI. As an AGI, it would be explicitly outside the scope of Open AIs license with Microsoft and must be made available for the benefit of the public at large. This is already a hell of a story with some pretty significant implications. One of the biggest ones we saw yesterday around the launch of Anthropics, Claude 3. As many pointed out, the presence of this lawsuit might significantly tie Open AIs hands in how fast they can move on new models. The lawyers may be taking over, limiting what Open AI can do, slowing things down generally,

and that indeed may be the point. This is definitely a story that's going to continue to evolve and one that I will keep watching here, but let's close on a song created by Riley Brown on Suno AI that he called Please Don't Sue Me Elon. Enjoy the track, and until next time, peace.