The AI Daily Brief: Artificial Intelligence News and Analysis - The Winners and Losers Following the OpenAI Power Struggle

Episode Date: November 24, 2023

As the dust settles, NLW looks at who has benefitted and who has lost out after Sam Altman's removal and reinstatement. Spoiler alert: no one comes out better. 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 the winners and losers in the Open AI fiasco. The AI breakdown is a daily podcast and video about the most important news and discussions in AI. Go to Breakdown.netnetwork for more information about our YouTube, our newsletter, and our Discord. Welcome back to the AI Breakdown. It is, of course, the Thanksgiving holiday here in America, and so I wanted to do a few episodes in advance to make sure that you guys had content over the long weekend. And one of the things that I wanted to do was take a step back and give a total. subjective opinion on the winners and losers in this whole OpenAI charade. Now, I will note that I am recording this on Tuesday night, and it is not even resolved yet whether Sam Altman and Greg

Brockman will ultimately go to Microsoft or whether they will stay at OpenAI. But I will argue that I think that we have enough of the story now to make some assessments about where things are going to land. And I'm interested to see if you agree with my assessments or not. One thing that I will note right at the beginning, is that I don't actually think anyone comes out as a clear winner. There are only varying degrees of losers and who lost the least. And at the top of the list of people who lost the least, let's say, is Microsoft. Now again, I don't know yet if Sam and Greg are actually going and bringing Open AI with them to Redmond or not, but Microsoft undoubtedly did rescue some amount of success from the jaws of defeat. It was an absolute masterstroke, too, after it had been

announced that Emmett Shear was taking over and that the board was appointing him as the new CEO, after Miramirati got all uppity and dying to bring back Greg and Sam, even though they had just promoted her. To have Satya just a few hours after that announcement, announce himself on Twitter that Greg and Sam would be coming to Microsoft, put them in a position where, at the very least, they stem the bleeding from what could have been a very painful Monday morning market open. That was sort of reinforced over the next couple days as the OpenAI staffers signed a letter on mass threatening to go to Microsoft with Sam and Greg, should the board not resign and reinstate them. Now, my strong guess is that this is not because more than 700 people on that team wanted to go work for Microsoft, but because that alternative seemed better than what the board was presenting.

So clearly, there are some benefits here for Microsoft. And it's likely to me that even if there is an arrangement that's figured out with Sam and Greg to come back to Open AI, that Microsoft will have more, not less control. There have been conversations around whether they'll get a board observer seat or even a board seat in some new reimagined structure. And certainly it's hard to imagine them having less control, especially if they ever sink in more resources. But why would they still be losers in all of this?

Well, two reasons. Let's assume that Sam and Greg do end up back at OpenAI. It will be an OpenAI that is weakened, that is slowed down, that has to do cleanup, that has to win back the trust of developers, an Open AI, in short, that is handicapped relative to the Open AI we had just about a week and a half ago. Microsoft's investment then is worth less, even if that's still the most desirable outcome. Now, what if the team comes to work at Microsoft? Well, there are some benefits for that.

One of the things that makes Microsoft's investment the riskiest was the fact that the OpenAI board had the ability to declare AGI, and Microsoft has no access to the IP. after that point has been reached, that is likely out the window if everyone comes and works at Microsoft. But frankly, even with OpenAI outside, Microsoft was already reaping the benefits to their core business. Microsoft's stock has been flying not because they were cool and smart to invest in OpenAI, but because OpenAI models are meaningfully increasing their Azure business right now. Relative to Google in particular, they outperformed market expectations this last quarter because of the growth of their cloud business, driven at least in part by their Open AI partnership.

In other words, they kind of have a have your cake and eat it two situation where they're making a ton of money for their core business with a startup investment without having to manage all those new employees or deal with the inevitable attrition that comes when startup people go inside big companies. So the TLDR on all of this is that even if it was a master stroke and a rescue to create an opportunity to bring that whole team over to Microsoft, I think that is still less preferable to just having OpenAI running as the startup they were. but even if that is the outcome, they won't be the startup they were. They will at least be temporarily handicapped relative to where they began. So Microsoft is the closest thing that you could have to a winner here, but they're still, I think, in the losing column.

Next up, OpenAI developers. I think they are more unambiguously in the losing column here. Why? Well, it's obvious. Working on OpenAI APIs is now a risk and a liability for more of the fact than they just might decide to compete with you. Now it has been proven and demonstrated that you have platform risk that STEM from their very weird organizational structure.

Already, some developers were nervous because it's not exactly clear which things OpenAI would ultimately want to just support through their APIs, and now this creates a whole additional dimension to that. Startups especially, any fiduciaries are going to be forced more or less now to diversify which APIs they use, which could be really challenging when OpenAI remains the state of the art in many cases. It just creates a new headache for an entire category of people who just want to be focused on building, and that is, of course, a net loss.

Now, to the extent that there is any upside, being less reliant on any single platform in the short term may save some heartache and pain in the long term, so maybe that forced diversification, even if suboptimal in some ways right now, will be net net a good thing in the long run. Next up, let's talk about the board as individuals. And by that I mean the remaining board, not Greg and Sam. They're clearly right now in the losing column. Why? There are some people who want to give them the benefit of the doubt and say they must have seen something terrible to make this decision. But the longer they don't articulate why, and what that is, the worst it's going to have to be to justify the chaos that has ensued from their actions.

Now, like I said, it's entirely possible that by the time you're hearing this, some shoe will have dropped that makes us all say, oh yeah, they were actually heroes. But right now, they look incredibly churlish. They look like they're dealing with old grudges, or just like they don't like Sam, and that they were willing to nuke billions of dollars in value for no apparent reason. This is only reinforced by the consensus with which the OpenAI team has rallied behind Sam and Greg and not behind the board. At the time of recording, something like 745 out of 770 employees had signed that letter demanding the board's resignation. Now, what's more, Adam DeAngelo specifically,

who appears to have become the ringleader, or maybe he always was, looks even worse because of the competitive dimension of this story, where Cora's Po chatbot was a direct competitor to both chat GPT and their GPT builder. And so people are, again, in the absence of any actual statements from Adam or other board members, assuming the simplest thing, which is the worst thing, which is that this was just petty jealousy and power struggle. Now let's talk about the board as an institution. They are firmly in the losing column. Why?

It's not because it didn't work to have a structure where even a very powerful CEO could be removed by a non-profit board. It did work. At least it appeared to have. The problem is that they kind of seem to have shot their shot way too early. Without justification and backup and really serious divergence that other people would agree with,

they are at least at the time of recording losing their battle. And so, this thing that was created to preserve the ability of a mission-oriented company to continue to build and do things, even though it had to have a for-profit entity to be able to raise the money that it needed to do those things, well, it actually came to pass. But by using that silver bullet too early, it mitigated the efficacy in a huge, huge way. Let's expand this more broadly to open AI governance in general. Because of the way that this has played out, a very interesting and potentially powerful experiment was made to look stupid and overcomplicated and subject to. frivolity and human frailty. What this likely means in practice is that rather than other companies in the AI space taking some interesting novel structure like this, or putting safety issues front and center in the way that they organize their companies, it's more likely that they're going

to go the opposite direction. More specifically, it's more likely that investors are going to demand they go in the opposite direction with various standard structures in corporate formats. Basically, the failure of OpenAI's governance here makes it less likely that novel governance experiments that could actually slow down the speed of AI if that's something that one cares about, will actually have a chance to breathe. And of course, that brings us to the next set that has lost huge from this, which I believe are the AI safety advocates. Now, at this stage, I should be very clear that it is not clear whether this was actually a fundamental disagreement around AI safety issues. In fact, did his note on Twitter accepting the OpenAI job, new CEO or current CEO at the time

of this recording again Tuesday night, Emmett Shear, the former CEO of Twitch, said that, it wasn't about disagreements about safety. But that doesn't matter when it comes to the narrative battles because this has created, or perhaps I should better say, supercharged a culture war. That culture war is, of course, between the effective altruists on the one hand and the EACCs or the accelerationist on the other. Now, EA has already been in a tough state ever since the downfall of Sam Bangman-F. People have been suspect of their motivations, their thought processes, and had questions about to what extent it was the EA-ness of SBF that led him to make the decisions that he made. Now we have a second Sam situation exactly a year later, and very quickly

Silicon Valley turned to the idea that this was some EA insurgency. The effective altruism forum has been lit up with comments around how this could end up being bad for EA ever since, and the acceleration as community in Silicon Valley in particular is totally emboldened. Now, we could have an entire episode around the space between two different extremes, but even getting outside of that, the fact that the board has said absolutely nothing about why they let Sam go, finds them to be more receptive to the arguments that many of the accelerationists and not even accelerationsists are making, that the AI safety folks are just crazy and out to lunch. Indeed, the ironic thing for AI safety advocates with this whole situation is that to the extent that it

was about preventing actions that were seen as against the cause of AI safety and responsible AGI, it's more likely that it has the exact opposite effect. A less public support for AI safety advocates, a Sam Altman that, whether in the context of Microsoft's new lab, or back at the helm of OpenAI, is going to be totally unbound by the constraints that had previously been set upon him, and potentially on top of that and related more concentration of power within Microsoft. And indeed, that gets me to the last group that has lost in all of this, which is all of us, humans. Why is that? Well, for all the reasons we just articulated, that this is likely to lead to more concentration at a smaller number of companies like Microsoft. The fact that a silver bullet

that some saw as important was shot too early. Or on the flip side, if you find yourself in the accelerationist camp, the fact that this potentially delayed progress. Point being that I can basically articulate how it's bad for everyone, even totally diametrically opposed people, but it's much harder to argue how it's good for them. Now, I suppose a potential counterweight was that open AI was in such a strong leadership position that having them knocked off, having developers have to use other things, for example, could be good in the sense that it limits monopoly potential. But then again, and even that comes at the cost of an accelerated AI arms race. And so where we end is basically that whatever happens, this was a very, very value-destructive event on numerous, numerous levels.

It was a value-destructive event from a capital perspective and from a market perspective. It was a value-destructive event in terms of the community that had been built around OpenAI, and that is now going to have to win back trust with developers. It's a value-destructive event in terms of any sort of reasonable participation that AI safety advocates had. It's a value destructive event probably in the eyes of Washington, which is something I didn't even get into. It's hard not to view it as fairly infuriating. Now, my hope is that in the subsequent few days between now when I'm recording this on Tuesday night and Friday when you're hearing it, that the situation will have settled down and resolved,

that all will have become clear, that the board will have at least tried to explain its actions in ways that even if we don't agree with, at least we understand better. But even with that, even if that stems some of the bleeding, I still think it all amounts to basically a damn shame. Anyways, let me know what you think I know you guys represent a huge array of perspectives from the AI safety side all the way up to the accelerationists. And while maybe nowhere else will be, this is a big tent for all of you. So let me know in the comments. Come join us on the Discord. And most of all, I hope you are having a wonderful gratitude-filled weekend.

Until next time, peace.