Let’s stipulate something: Making fun of AI chatbots is cheap and easy. But it’s also a little fun. And, frankly, I was a little curious which color Sonos Ace headphones ChatGPT might recommend.
So I asked it.
Specifically, I hit up ChatGPT 4o. That’s the version “that provides GPT-4-level intelligence, but is much faster and improves on its capabilities across text, voice, and vision.” One model is generally as good (or bad) as the next as far as I’m concerned. I wouldn’t take any of them as gospel. So consider this an exercise for entertainment purposes as much as anything.
The first question I posed: “What color Sonos Ace should I get?”
Before I get to the reply, I just want to say that I’d consider that to be a fairly natural way to ask that question, right? Presumably, ChatGPT would be able to suss out that the Sonos Ace are a thing (albeit a relatively new product, which may actually be problematic for the response), and that they come in multiple colors. And it did both of those things.
But here’s the problem. Two of them, actually. First is that ChatGPT didn’t understand that the
That first part led to a pretty funny response. “The choice of color for a
And it gets funnier. Here’s the response for
But what about white
Then ChatGPT offered up a third option. “Neutral and versatile, gray can be a good middle ground if you want something that isn’t too stark but still stylish and modern.” Makes sense. But there’s just one problem:
We’re not done yet. ChatGPT then mused on “special editions or custom colors.” Fair enough. “Occasionally,
Again, this was really all done just out of sheer curiosity. I’m not looking to make ChatGPT feel bad or anything. But the idea that it didn’t recognize on its own that the Sonos Ace are headphones sort of gave the impression that it doesn’t actually know what they are and was just giving me responses it thought would be somewhere in the ballpark, which really just made them generically bad. Like a hastily written paper in sixth grade.
I ran my query once more, because I’m a glutton for punishment. This time I added in an extra word, for context: “What color
The responses were better in that they didn’t assume the
- Pros: Classic, versatile, and less likely to show dirt or wear. Matches well with most outfits and styles.
- Cons: May feel too plain if you prefer a more vibrant look.
Again, generic. But not wrong. The answer for the white
But, again, it also dreamed up a gray edition of the
The point of all this? I just wanted to see how a newfangled AI model answered a question I’d already answered myself.
Phil 1. ChatGPT 0.