Ask Paul: August 7 ⭐

Happy Friday! And welcome to another epic installment of Ask Paul after a very quick trip to Las Vegas, with some terrific reader questions and some overly long answers. What can you do?
"Moving the Microsoft Edge extensions ecosystem forward with Manifest Version 3"

Ironically, AI found a typo there: You mean "backward"

🔥 Hot or not?
NickTech asks:

I saw your article Comedy or Tragedy? regarding Microsoft financials around the same time I saw other articles about Microsoft's profits in AI likely being 70 percent from OpenAI.

To be clear, these are revenues (sales), not profits (money left over after you figure in the cost of doing business). Microsoft's AI push will never be profitable until the company writes off or otherwise closes the books on the "investments" it's now making and can start fresh from a financial perspective.

With Microsoft releasing their own AI models, it seems like they’re at least trying to be semi-independent from OpenAI in that field, possibly even fully independent in the future.

Correct. That is the goal: To lessen the reliance on OpenAI as much as possible and as quickly as possible. Of course, growing a family of frontier models from the ground up is the one thing that doesn't happen quickly with AI. So this will take a while, which is why I was amused to see headlines about Microsoft starting to use its own models for core Microsoft 365 features. This was a baby step, important, yes, but not a major advance.

Considering Microsoft’s history of not being able to break into other markets, do you think they can succeed in AI, or have they missed out on this again?

They will succeed. With the caveat that there are multiple angles to this.

There's the consumer-facing, heavily publicized AI, the AI assistants that everyone knows and some seem to love. There's the heavy lifting that needs to happen on the backend, the infrastructure parts that only the world's very biggest companies--Amazon, Google, Microsoft, and a few others--can afford to bankroll. And then there are these numerous side efforts that scale up and down the spectrums of cost and scalability that include such things as Chinese models, open weight models, second- and third-tier models, and local, on-device models, the latter of which are arguably the most important for making AI affordable and available to all.

Microsoft will never be successful with consumer AI. We can just write that off. But that doesn't mean that the Microsoft products and services that deliver AI features won't be successful: We will use Windows, Office, OneDrive, and whatever else, and we will have access to some levels of AI functionality depending on what it is we pay for or don't. It's just that consumers will not pay for Copilot as a standalone service. Ever. I will die on that hill.

Microsoft will have some level of success with AI for businesses because Microsoft has tremendous reach with businesses across the Fortune 500 through Microsoft...