Microsoft Sends a Signal: AI, AIn’t Working
March 11, 2025
Another post from the dinobaby. Alas, no smart software used for this essay.
The problems with Microsoft’s AI push were evident from the start of its AI push in 2023. The company thought it had identified the next big thing and had the big fish on the line. Now the work was easy. Just reel in the dough.
Has it worked out for Microsoft? We know that big companies often have difficulty innovating. The enervating white board sessions which seek to answer the question, “Do we build it or buy it?” usually give way to: [a] Let’s lock it up somehow or [b] Let’s steal it because it won’t take our folks too long to knock out a me-too.
Microsoft sent a fairly loud beep-beep-beep when it began to cut back on its dependence on OpenAI. Not long ago, Microsoft trimmed some of its crazy spending for AI. Now we have the allegedly accurate information in “Microsoft Is Reportedly Potting a Future without OpenAI.”
The write up states:
Microsoft has poured over $13 billion into the AI firm since 2019, but now it wants more control over its own models and costs. Simple enough in theory—build in-house alternatives, cut expenses, and call the shots.
Is this a surprise? No, I think it is just one more beep added to the already emitted beep-beep-beep.
Here’s my take:
- Narrowly focused smart software adds some useful capabilities to what I would call workflow enhancement. The narrow focus for an AI system reduces some of the wonkiness of the output. Therefore, certain tasks benefit; for example, grinding through data for a chemistry application or providing a call center operation with a good enough solution to rising costs. Broad use cases are more problematic.
- Humans who rely on information for a living don’t want to be caught out. This means that using smart software is an assist or a supplement. This is like an older person using a cane when walking on a senior citizens adventure tour.
- Productizing a broad use case for smart software is expensive and prone to the sort of failure rate associated with a new product or service. A good example is a self driving auto with collision avoidance. Would you stand in front of such a vehicle confident in the smart software’s ability to not run over you? I wouldn’t.
What’s happening at Microsoft is a reasonably predictable and understandable approach. The company wants to hedge its bets since big bucks are flowing out, not in. The firm thinks it has enough smarts to do a better job even though in my opinion this is unlikely. Remember Bob, Clippy, and Windows updates? I do.
Also, small teams believe their approach will be a winner. Big companies believe their people can row that boat faster than anyone else. I know from personal experience and observation that this is not true. But the appearance of effort and the illusion of high value work encourages the approach.
Plus, the idea that a “leadership team” can manage innovation is a powerful one. Microsoft’s leadership believes in its leadership. That’s why the company is a leader. (I love this logic.)
Net net: My hunch is that Microsoft’s AI push is a disappointment. Now the company can shift into SWAT team mode and overwhelm the problem: AI that does not pay for itself.
Will this approach work? Nope, the outcome will be good enough. That is a bit more than one can say about Apple intelligence: Seriously out of step with the Softies.
Stephen E Arnold, March 11, 2025
Comments
Got something to say?
-
- Subscribe to Beyond Search
Feature archive
News archive
-
