There are moments when you realize technology has advanced too far. Self-driving cars. AI girlfriends. Refrigerators texting you about lettuce. The latest is an AI-powered collar that claims to translate canine noises into human speech.
Naturally, I bought one immediately.
When my wife returned home after I outfitted our dog with PettiChat, Zoe barked and jumped up to greet her, as usual.
“Zoe’s so bouncy, glued right to Jo Ann,” the collar said in a young female human voice.
My wife froze.
“How the f***?!” she demanded to know.
During the setup process, the PettiChat app asked me to characterize our doggo’s default attitude. It offered eight choices, from happy to mischievous to sad. I chose “clingy,” because she herds us like she’s auditioning for the next Yellowstone spin-off, freaking out if my wife or daughter go off on their own. Me, not so much.
The app also asked me to input the dog’s sex and the first name of the human she’s most attached to.
“I don’t know how!” I replied to my wife. “That’s crazy, isn’t it?!”
Unfortunately, this was the most fun we had with this experiment, and it was over in five minutes—the same amount of time it took my wife to figure it out and break me.
Much Ado About Ruffing
A Chinese AI tech startup called Meng Xiaoyi built PettiChat by combining motion detection and a microphone with Bluetooth access to an AI database. Alibaba Cloud Qwen is a large language model supposedly trained on millions of animal recordings.
It claims that PettiChat “lets your dog talk to you with 95% accuracy.”
If that were true, we most certainly wouldn’t still be cleaning up poop from the carpet underneath our bed. And Zoe would have asked us at least once, “Why the f*** do I always have to eat the bad food while you eat the good food?”
The manufacturer reported more than 10,000 preorders at $150, because people desperately want to communicate with their fur babies. Also, because they’re stupid.
Since Zoe doesn’t bark much during her default state—sucking the heat from the electric blankets on our sofa—PettiChat was most effective to us as a training tool. Basically, it helped us train her to be annoyed and bark more, which is exactly what you don’t want in a dog.
All three of us couldn’t help trying constantly to make her “speak” to test whether the collar could guess what she was saying. We clapped. We danced. We pretended to fight. We did all the things Zoe does not tolerate in our apartment.
“Rarr! Zoe doesn’t wanna do that, Jo Ann, let go please,” PettiChat said after I pretend smacked my wife’s face, and she recoiled backward in make-believe pain.
“It, like, low-key works!” our daughter explained in her 15-year-old way while hardly looking up from TikTok. “That’s scaring me.”
Basically, though, PettiChat didn’t tell us anything we hadn’t already figured out ourselves for free. Then it guessed—based on the info I had previously input—like a dime-store psychic on the rest.
“Pummped, Jo Ann—Zoe wants to snuggle close and look at that new thingy!”
“Zoe notices that odd, little thing nearby, so curious!”
“Zoe spots tiny flickers, wants to dash right away, Jo Ann!”
And it was flat-out wrong a lot more than 5%. For example, when it declared, “Brrah! Zoe is so mad—get that annoying thing away right now!” when she barked because our daughter’s best friend came over. She loves Ava.
Maybe the collar is better with cat meows, with which it claims the same 95% accuracy rate.
Oh, it did get one thing repeatedly correct: PettiChat helped Zoe tell us—on several occasions—that our WiFi was disconnected.
Such a smart dog!
