Understanding the Microsoft and OpenAI Relationship

Disclaimer - this is all educated opinion based on observations and decades of experience working with Microsoft and its employees, tools and servers.

Sam Altman going to Microsoft is a game changer. Here's why!

If you unplugged on Friday to take the weekend off from all the news, you missed quite a story that has unfolded until Monday!

- Sam Altman was fired from OpenAI

- Greg Brockman, President of OpenAI leaves with "If Sam goes, I go"

- OpenAI names interim CEO

- Interim CEO says they are bringing Sam back as CEO

- Board announces new CEO...and it isn't Sam

- Microsoft announces "Oh, yeah, by the way, we've hired both Sam and Greg for a new AI group here"

It can seem sort of chaotic and haphazard, but I assure you it isn't.

I have been in the Microsoft development space for over 30 years, building software, using their servers, and hobnobbing with their employees in the development space. I have watched AI for all that time and was fascinated by OpenAI when they exploded into the mainstream in Nov 2022.

I watched as Microsoft invested in OpenAI and started to use GPT in its own offerings, hinting that ChatGPT would be integrated into everything. It was a match made in heaven -- the dev and application company bringing in the AI technology of an AI leader, giving that AI company a huge user base. What a win-win!

So, the news today of Sam Altman and Greg Brockman joining Microsoft might seem odd, opportunistic, and exciting, depending on your angle.

I mean, does them joining Microsoft strain the expensive relationship between the two companies? Does OpenAI have any means to push back on "Microsoft's" (Sam and Greg's cough cough) recommendations?

Let's think about this a little:

• Microsoft invested $10 billion into OpenAI
• Microsoft quietly dismantled their own AI group completely after slowly reducing its size and impact. This gave clear AI control over to OpenAI. At first I thought it was just a recognition that an internal Microsoft team would be redundant and unnecessary. Today, I am thinking it was clearing the decks to make way for those two to come on board - if not in a plan, then in a contingency.
• Microsoft is betting the bank on copilots, integrating them into Office 365 and Windows, touting them as how things will work going forward. As a way of integrating AI into day-to-day tasks. Copilots are basically chatbots that do tasks for you too.

• Microsoft claimed it was GPT-4 in their copilots as they rolled out the idea to us devs. It was exciting! Microsoft was on 4, while OpenAI was still on GPT-3.5-turbo. It felt like Microsoft was getting the best of OpenAI's offerings and helping to define direction for the model. My thought at the time was that Microsoft was getting the cool first-fruits of the AI stuff. It pointed eyes at Microsoft as an AI leader. Well played, to be honest.

• Over time, however, Microsoft started saying "It isn't REALLY GPT-4 in our copilots but our own version of it." So did they take the GPT-4 model and start bastardizing it for the tasks the copilots would be doing? Makes some sense...take the "generic" GPT-4 model and train it with Microsoft's world of apps and programs. This would make sense. So now you have GPT-4 and MicrosoftGPT-4. Begs the question of what happens with OpenAI goes to GPT-5? Or GPT-4b? When the core model changes, how will Microsoft work it into their bastardized model? (Remember, MS had no AI group any longer)

• Copilots have struggled to be adopted so far, and there is good reason for that: they are clunky, noisy, and overreaching. They take up a lot of screen real estate without offering real help. And when you ask for help, it churns before spitting out an ok answer but not really what you want. I think copilots are the way to go, but I think the Microsoft versions to date are typical Microsoft trying to be something to everyone. Anyone who has gone through Microsoft Bob, the Microsoft Phone, Windows 8 and others can see the writing on the wall if copilots don't shape up quickly. (Targeted baby steps will be the way to win here)
• OpenAI starts heading into apps instead of advancing only the GPT model. Not a bad thing, but they were never an app company and Microsoft is built around development products.
• And what does this say but there is a bit of a rift between what OpenAI is offering and what Microsoft wants out of the model! The path forward has to have had some interesting shouting meetings.
• As a Microsoft environment dev, and a guy building plugins and using the OpenAI APIs, I can say OpenAI's plugins felt like a strained set of offerings, and sort of hokey. More like samples of what could be done with the underlying model than ever being a thriving ecosystem. The store was hard to use, it had all sorts of crap on it, and they never tried to make it a real store. That felt purposeful to me.
• Microsoft has development down cold and wanted the AI in it. OpenAI has AI down cold and wanted to build dev tools that use it.
• Keep in mind that ChatGPT is just an application. It is a website on top of the GPT-x models. Plugins made sense for helping developers use the GPT models and produce something visible via the ChatGPT ecosystem. It is clear OpenAI embraced a very simple ecosystem model and never intended for it to be a "thing".
• OpenAI bringing out GPTs, custom ChatGPT basically, started them more fully down a split path of model builder and creator tool builder. That split caused a split in OpenAI that is widely reported. Some purists really want OpenAI to focus only on the model part, on some of the issues that keep coming up in the models. What about governance, and security, and hallucinating and truthful results and lack of bias, and, and and...
• In typical Microsoft fashion, they continue to try to push technology on people by forcing a specific look and feel and result from their copilots. Personally, it is clear they are trying to use marketing and app ideas to make a development and technical thing like GPT "available and accessible" to the masses. (Stylish and usable apps are not their strong point...their strength is in the TOOLS, not the apps themselves)
• So the last couple of months show rifts in BOTH companies as they deal with their longer-range visions of where each company should be, and how their dance together should go.
• Without an AI group, and with a growing rift within and between the companies, there was one very good solution staring everyone in the face: bring on the best of OpenAI vision, enabling them to make a true ecosystem of ideas, and build out MicrosoftGPT from there.
• Altman and Boardman are free to let their vision go wild.
• They bring their GPT-5 ideas with them and can begin building a huge AI team to make them happen.
• Microsoft can continue the OpenAI investment so that OpenAI can retool and focus on their strengths. But I expect that it will be a sort of afterthought investment. See it through and don't renew it.
• Microsoft will build up their AI team from a new foundation. One that they paved the way for. They saw good underpinnings in plugins and APIs and GPTs...something they could expand upon into their own development tools.
• If Altman and Boardman are given the freedom, power and authority to make "OpenAI 2.0: Microsoft Land" a reality, then the model and functionality have a lot of potential. If Microsoft tries to control them, copilots and AI from Microsoft will die a quick death and we will see the two new hires moving on to their other ventures waiting in the wings. My gut from past experience says it will be the latter, but I do hope I am wrong.
• OpenAI will struggle. But they were well on their way with executing a vision and two guys don't make the whole company. There are good people there that can re-focus and make things shine. But I don't see them being as relevant without Altman/Boardman's guidance.

So where does it leave us?

So where does it leave us? You know, the people who rely on ChatGPT and the GPT APIs for so much commerce and help and tooling and and and?

For me personally, I am staying the course. I use ChatGPT, build plugins and GPTs, and assume they will figure it out. But instead of keeping one eye open, I now have both eyes open.

None of Claude, Bard, and the other models are ready to take the throne. If ChatGPT goes away or goes downhill, it will set AI back quite a bit for some time. Not in the field, but in the eye of the common person and companies. The other models are improving but have issues still, beyond what we see in ChatGPT. People will lose momentum. Lose excitement. It is already happening with the current issues in GPT-4. If that model takes a big hit or goes away, then it won't be pretty. Trust is hard to re-acquire.

I fully believe that the ball is in Microsoft's court, not OpenAI's. If they make MicrosoftGPT great, then they have the means to put it into developer hands (both Microsoft environment and not) who then bring it to the world. If they try to control it to force copilots to be the focus, then it dies.

MicrosoftGPT is not a real thing, just a name I give it.

But if you'd like to have my help building against GPT or MicrosoftGPT with custom apps and chatbots, then DM me or contact me at [email protected].

I also have a program called "The AI-Enabled Solopreneur" that helps solopreneurs find the right spot for AI in their business and execute a well laid out plan to free up their time, scale their business and reduce mundane tasks. We are booking into 2024, so DM me for details on which cohort is available for you and for more information on this transformative program!