OpenAIs Problems Are Getting Worse....

So are things getting worse for OpenAI? Recently there was a story about how Microsoft are essentially having a bunch of issues with OpenAI and apparently Microsoft may actually want to get a little bit more independence from OpenAI rather than relying on their AI models. Let's actually take a look at this story because it goes a little bit deeper than you do think. So let's take a look at this article, which is the one that was really, really intriguing. And this was the article that had a lot of people freaking out about OpenAI and Microsoft's. Now, essentially one of the big points of contention was the fact that Mustafi Sullan wasn't getting the answers that he wanted. Apparently there was a recent video conversation which is of course basically a zoom call in which Mustafa Suleiman wasn't getting the answers he wanted. Now, if you don't know already, Mustafa Suleiman, he's actually the CEO at Microsoft's AI department. So this is the guy that is leading the AI section at Microsoft. And of course these individuals who are working at the Microsoft AI section, they are really trying to work with OpenAI because of course have some partnership now, of course it says here that last fall during a video call with senior leaders at OpenAI and Microsoft, Mustafa Suleiman, who leads Microsoft's in house AI unit, wanted OpenAI staffers to actually explain how its latest Model 01 worked, according to someone present for the conversation and two other Microsoft employees. But he was actually peeved that OpenAI didn't provide Microsoft with the documentation about how it had programmed O1 to think about users queries before answering them. And I think the reason that OpenAI have done this is because they want to keep that chain of thought to be as proprietary as possible. Of course there are different methods. I think everyone knows that the model thinks before responding, but they want to essentially of course figure out how O1 works behind the scenes so that they could potentially replicate that themselves. Now I'm not sure on the agreements between OpenAI and Microsoft, but I don't think it was written in that OpenAI would reveal all of the secrets to their models. And clearly it's, you know, shown that Microsoft were really frustrated by this, or should I say Mustafa Suleiman. Now the crazy thing about this is that I can understand why OpenAI wouldn't want to reveal those secrets because of course those secrets are essentially at least the key to their moat. So the fact that they wouldn't want to share them doesn't make sense because otherwise Microsoft wouldn't really need them. And of course that information could proliferate throughout the AI community, resulting in a diminished lead for OpenAI considering the fact that many companies are circling around the generative AI space, which is of course something that OpenAI are trying to keep the lead on. Now if we take a look at what Microsoft have already done, I'm not sure you guys might remember this, but Microsoft have their own model which, which is MA1 and this is where Microsoft are readying a new AI model to compete with Google. So if you remember, Microsoft have been internally working on a secret model because they wanted a model that could perform as well as ChatGPT so that they would have less reliance. And this was something that they spoke about, I think this was a couple of months ago. So if you actually take a look at this, you can see here the model was called MER A1, 500B and this was supposed to be a larger model which is of course larger than the open source models that Microsoft previously trained. That is of course relating to the Phi series, which is a series of really small models that Microsoft trained that are looking to go on device for many different use cases. And the FI series is actually a really great series because it is roughly the size of 2B 7B and I think 14B. And those models have comparative performance to models like ChatGPT. But with the MAI1 they actually wanted this model to be basically like GPT4 and of course they wanted this model to be really good. But you can see that OpenAI weren't willing to reveal some of the secrets that they actually used to get to this level. So you can see right here it says, raising his Voice, Suleiman told OpenAI employees, including Mira Moratti, then OpenAI CTO, that the startup wasn't holding up its end of the wide ranging deal that it had with Microsoft. And then the call ended abruptly. So he was like, look, you guys aren't really giving out your information, but what's going on here? And you know the call ended abruptly. So that is of course not something great. And you know the article continues to talk about how there is conflicting mandates according to executives in his unit. So first he must simultaneously carry on the OpenAI partnership through which the two companies share information about AI research and development. But at the same time, he also is under orders to put Microsoft on a path to self sufficiency in AI, so it won't have to rely indefinitely on OpenAI's technology for the majority of Microsoft's AI products. So you can see right here, it's pretty difficult to do that because they're conflicting interests. If you're working with OpenAI on AI, of course you're going to be talking with them about a lot of AI research. And if you talk about certain pieces of AI research, which are breakthrough pieces, then you could of course use that knowledge and information to power Microsoft's AI products. But of course these two companies on the model side, then of course they would be competing. So OpenAI probably thinking, you know what, it probably doesn't make sense, but all of this isn't really good because one of the problems, and I'm guessing this is why Microsoft is kind of frustrated with OpenAI, is the fact that, you know, the recently completed training of the family of Microsoft's models, internally referred to as mai, that perform nearly as well as the other models from OpenAI. This was something that didn't actually work and I will talk about that in a moment. But they did recently complete training of this and remember they actually spoke about this last year, I think it was mid last year. And now we're in 2025 when they were training those models. And apparently they're also training reasoning models which use a chain of thought. And of course this could compete directly with OpenAI, which once again is what I'm talking about. So it's not really great that OpenAI is facing such competition from a company that is working closely with it. You can also see here that Microsoft staff already experimenting with swapping out the Microsoft models, which are far larger than an earlier family of Microsoft models called Five for OpenAI's models in Microsoft's Copilot. And the company is considering releasing these models later this year as an API, a software hook that would allow people to weave in these models to their own apps. So there's a potential for these models to be released as an API later on this year, which would be super, super intriguing. And of course, at Suleyman's directions, Microsoft has been hedging its bets further by trying out models from OpenAI's competitors to even Power Copilot. And you can see right here that one of the craziest things about this as well is that Microsoft's also been hedging its bets actually look at different models to try out copilot. So I don't know if this is because they've been arguing or there have been some disagreements, but Microsoft looking to get some models like deepseek. And those are the models you can see right here. It says from Anthropic, from Musk's AI along with open source models from deepseek and Meta platforms. This is what they are looking into to see how those can work when powering copilot. And I honestly think that maybe people wouldn't tell that much. I don't know about you guys, but anytime I've used Copilot it's been very rudimentary and, and I don't think the model needs to be that good in order to achieve its chance. That's maybe just me because maybe I don't use it that much, but I would presume that most cases, most use cases for copilot are pretty timid. So if you do swap out the model, maybe it might not impact user engagement that much. That is of course not really a good look for OpenAI. Of course you do want Microsoft to be using your models, kind of dependent on your models, so you can definitely build them for the API essentially. So you can see right here that it says it's an incredibly competitive and creative time. Suleiman said in an interview you we are now using under the hood essentially all of the models from major labs, including all of the open source models and we are experimenting with them, we're flighting them and that's not what anyone thought would happen. So it seems like, look, they are no longer just depending on OpenAI because of course if they're not going to be like working with OpenAI in terms of getting that kind of data from them so that they can power their own ones and since they're MA1, there are actually some failures, it makes sense for them to kind of look out elsewhere to see what kind of other models they can use for copilot. And I think it's a situation where maybe the ChatGPT API is considerably more expensive than other free alternatives like deepseek. You can see right here as well. Like I said before Mai1 I'm just going to call it my one or the my models took nearly a year due to technical setbacks, so abrupt changes in strategy and the departures of several Microsoft veterans who disagreed with Suleyman's management and technical approach, according to interviews with six current and former Microsoft employees. So, so maybe it's painting the picture here that maybe Suleiman was a bit of a disagreeable person because it's saying that six people disagreed with his technical approach and it's saying that during that time, while Microsoft were kind of stuck with this one model, this my model, OpenAI had already gone and published several batches of cutting edge models and those setbacks have left some outsiders unconvinced that Suleyman can deliver on his grand strategy. So this is something that internally, Microsoft, it seems like things aren't going that well, especially with them trying to figure out what they're going to do with their AI models. And, and the reason that I said this is not good for OpenAI is because if they reduce their reliance on OpenAI, then of course OpenAI won't really be serving them anymore. It says here from the outside, it's still not clear what they have to show for the creation of this unit under Suleiman a year later. This is what a venture capitalist is saying. And what's crazy is that at the time the models were actually stated of performance, but they declined to specify how they were going to use those models in practice. And if we look back at this interview, this was an interview with Sati Nadella on the Dwarkrish Patel podcast. And this was something that was really interesting because one of the things that, you know, that was actually spoken about in this episode was the fact that Microsoft are going to be using open source models and the fact that they are going to be looking at a variety of different things, not just relying on OpenAI. And they spoke about how the model space is changing and OpenAI's dominance won't necessarily always be there.

So that I think is what will happen even in the model side. So there will be open source, there will be a governor, just like on Windows. One of the big lessons learned for me was if you have an closed source operating system, there will be a complement to it which will be open source. And so to some degree that's a real check on what happens. And so I think in models there is one dimension of maybe there will be a few closed source, there will definitely be an open source alternative and the open source alternative will actually make sure the closed source winner take call is mitigated. So that's kind of at least my feeling on the model side.

And then even in the CNBC interview which I showed at the beginning, you can see that there were some not claims, but some questions being put to Satya Nadella about the OpenAI relationship being a bit rocky at the moment and seeing where things are heading from there.

We'Re going to go off longer term and build your own frontier models and that there's a disconnect or of some sort between Sam Altman and Mustafa, who you hired for to run your AI program. Can you speak to that? Look, I think the partnership with OpenAI to us is a critical partnership. We love it, it's working, it's created a lot of value for us and we plan to just continue forward with it. It's true that on Azure today there are other models, there are open source models, we have our own small language models and we'll continue to have all of that because customers at the end of the day are going to demand choice there, and that's fine, and we'll make sure we support it.

Let me know what you guys think about that. Is it the end of OpenAI Micro? Microsoft.