“Hey Google, how are the Giants doing this year?”
I was standing at my bathroom sink, finishing up my daily ablutions, when a random thought popped into my head. It’s been a minute since I checked in on my favorite baseball team – ever since I moved East, they’ve been in a slump. Maybe they’re pulling back into contention this year? I still have a Google Home plugged in nearby. I’m intimately familiar with the Home’s limitations, but I asked anyway. Perhaps I was hoping one of my annual predictions (voice interfaces for the home) would magically come true.
Nope.
“The San Francisco Giants are in fourth place in the NL West.”
In a tech world driven by generative AI, that’s a pretty lame answer. So I headed into my study and asked Google on my computer. Better: A search one box showing the Giants’ place in the NL West (it’s not pretty). Sure, there were the traditional blue links below, but that one box gives me a better sense of the answer.
Still no AI magic, however. I pulled out my phone and asked Google in “AI Mode” – a habit I’m not yet in, but one Google is pushing through incorporating AI into its core search offerings. Now I get something quite different. A chatty AI agent summarizes any number of relevant angles on my team’s dismal performance:
Why am I on about this? Well, I’ve been mulling the role that basic economics plays in the unfolding story of AI in our daily lives, and I’ve come to realize that the magic of AI isn’t coming to save us any time soon, at least not when it comes to what pundits like to call “ambient intelligence.”
This all relates to what feels like a significant vibe shift in the AI landscape. Three-plus years into what was sold to us as a revolution, AI remains mostly tethered to its original form factor: a helpful if somewhat frustrating chatbot that lives on your phone or web browser. Yes, we’re finding more and more things to do with this new pal of ours, but for many of us, AI is losing its magic.
Why? Blame economics. Serving up AI answers remains an extremely expensive endeavor. Rationally, firms with the wherewithal to deliver AI are productizing their services in places they believe will provide the maximum opportunity to deliver a return. For Google, that’s the search box and their Google Workspace surfaces. For Meta, it’s engagement bait to keep people on Instagram and expansion into business markets. Over at Microsoft, it’s all about enterprise applications. Amazon? Commerce, of course. And Apple? Still waiting it out, though it seems they’re on the brink of a better Siri.
In other words, so far it’s been more of the same. And more of the same isn’t exactly magical. In fact, it’s starting to feel like we’re being steered, and we kind of don’t like that.
Claude is my AI chatbot of choice, so I asked it the same question. Its answer:
Not good, John — they’re 23-38, dead last in the NL West and essentially tied for the worst record in the National League (only the Angels are clearly below them; the Rockies are a hair ahead at 24-38). They’re sitting roughly 16 games back of the Dodgers in the division, which at this point in June is less a deficit than a verdict.
The recent stretch tells the story in miniature. They hung 19 runs on Colorado on May 31 — which sounds great until you remember Colorado is a 24-38 team and Coors Field flatters everyone — and then ran straight into Milwaukee, the best team in the NL at 37-21, and got outscored 24-5 across the first two games of the series. That’s the pattern of a bad team: they can beat the other bad teams and get steamrolled by anyone good. A 19-run outburst against the Rockies doesn’t tell you much; getting held to 2 and 3 runs by a real pitching staff does.
On prospects for the next couple of months, here’s the honest read, and I’ll flag where I’m reasoning versus where I actually know something. The schedule ahead is survivable but not soft: they finish at Milwaukee, then a four-game set at the Cubs (32-29, a legit team), then home against Washington (a sub-.500 club they should split or win). So there’s a chance to pad the record against Washington, but the Brewers and Cubs portions are likely to hurt. None of that changes the structural problem, which is that a team 15-plus games out in early June has, historically, essentially no realistic path back into contention — the math of needing to play .650 ball for four months when you’ve played .377 for two just doesn’t bend.
Now that’s a damn fine answer, and I’m paying $100 a month to get it (I’m on Claude’s “Max” plan). But wouldn’t it be cool if I got that kind of a response on my Google Home? After all, that’s how this whole thing started, back when I was brushing my teeth at my bathroom sink. I have a notion, I speak it into the air, and the room fills with an intelligent and engaging answer. I’ve essentially summoned a seasoned baseball color commentator into my bedroom, and I can speak with him with all the fluency of Captain Kirk on the bridge of the Enterprise.
Computer! What’s wrong with the Giants?!
Well John, how much time do you have?!
Magic!
Well, that ain’t gonna happen. First, Google is never going to export its Gemini service to its Home business, because the economics simply don’t work. There’s no ad surface to speak of, and no one is going to pay a subscription fee for a home speaker, especially if they’re already paying for Google Workspace, OpenAI, or Anthropic’s AI offerings.
And while it’d be cool, I can’t imagine Anthropic or OpenAI getting into the home speaker business. Anthropic makes most of its money helping enterprises code, and OpenAI is on a path to becoming an advertising business. Again, it’s hard to build a viable ads business on top of a home audio device.
I’m belaboring this example to make a simple point: The idea of “ambient intelligence” is super cool in theory, but dead on arrival in practice. And that fact won’t change until the economics of delivering AI change by several orders of magnitude. I’m still six months from grading my 2026 Predictions, but I’m pretty sure I got this one wrong.
–
so spot on and that horrible reminder that the ad-driven web legacy hog ties so much future tech that might have been embraced and useful to a wide range of people