In 1948, Cadillac introduced their first car with tailfins. Arguably some other cars may have had tailfins prior to that car, but the tailfin war is pretty much agreed to have begun with that Cadillac and its tiny little tailfins.

Over the next 20 years, fins would grow, shift, move sideways, and become a bizarre design element across most of the auto industry, Even Mercedes-Benz wasn’t immune to this design trend, featuring “fintails” or “peilstege” until 1968 on four-door sedans, at a time when most other automakers had already abandoned them.

While most consider tailfins to be a design element, some considered them a safety feature. Particularly on land-yachts of the era, tailfins could arguably help when parking the vehicle, by helping to establish the outer perimeter of the car at a time when cameras and parking technologies were not even a dream on the horizon (M-B’s “peilstege” term was a specific example, as it meant “bearing bar”, a feature of the car to help you get the bearings of where the car ended.)

However, tailfins weren’t a safety feature. Because of their design, they made accidents worse for those who were hit and those who hit them. They even led to increased pedestrian injuries when sitting still.

By 1996, when Cadillac finally dropped their Fleetwood line, even vestigial tailfins were finally dead. Although I personally feel like they still haunt Cadillac design today, for no good reason other than being a weird legacy design element of Cadillac’s overall design language… or something like that.

So why ramble on about tailfins, when the title of this post talks about AI?

Because I maintain that AI (and really machine learning as a whole) are tailfins. “AI” is the new “space age”, the new “plastics”, the new “must have” that technology purveyors are convinced will land them immense revenue that they’ll miss out on if they don’t also keep up with the joneses and add ridiculous, minimally functional tailfins to their products.

I recently noticed that a rice cooker I have and love that uses “fuzzy logic” now simply refers to it as “AI” on newer versions.

That’s stupid. Don’t fall for that.

But this isn’t about that. This is about Microsoft, Google, and Amazon all defocusing so much on “AI” that they are failing to feed and water their bread and butter products and services. Instead of addressing customer needs, they are spinning up infrastructure and stupid amounts of energy inefficient Nvidia GPUs to try and win at a gold rush… where there is no gold.

Most normal people don’t care about AI. Many of us don’t even want it. But for the love of all that is good, if you’re going to add “AI” slop to your existing products and services, don’t do it at their expense. Look at Copilot in Windows. A shadow of the weird initial vision Microsoft had for it (change your dark/light settings with a simple command!), it’s now just a wrapper around the version of Copilot within Bing. Microsoft talked over a year ago about Copilots other than GitHub, but has seemingly struggled to deliver a cohesive set of technologies that actually make sense, and their branding struggles have been even worse. Worst yet, while Apple’s “belated” Apple Intelligence technology knows when to tell you it can or can’t do a task when offline, because Microsoft has such an unfulfilled on-device “AI” strategy, they can’t do almost anything offline on a modern PC.

Microsoft in particular has me perplexed…

• They’ve sunk so much cash into OpenAI/ChatGPT and the hardware and infrastructure to run it that they’re pretty clearly willing to burn the ships to try and make this fetch strategy happen
• But their approach is so sloppy, marketing so unclear, and feature delivery so inconsistent, that it’s hard to tell how committed product teams really are to this grand vision.

I kind of get the promise of AI. I can also see the theory (as yet unfulfilled and unfulfillable) that “AI” could lower labor costs – by saving time for individuals on their own work, and theoretically saving the amount of individuals an organization needs for their workforce, by replacing them with “AI”.

But I believe in this space, investing in robotic process automation (RPA) is more logical than investing more deeply in “AI”… the unknowns, and worse, the knowns (cost, heat, water, electricity) about “AI” are too big to ignore. RPA at least lets you wrap your existing business processes in existing technology, with little to no “magic” involved.

But the automakers are telling us we need bigger and bigger tailfins, without showing evidence that these tailfins actually do anything positive… and failing to show how they are offsetting the practical environmental damage that “AI” is already threatening. Sure, they will tell you that with just a few billion more, and a few more iterations of software, that it’ll all be perfect.

But a billion here, a billion there, eventually you’re talking Windows Phone money.

??