Interesting that they think there’s a single universal voice amongst hens. I’ve got a small flock of a dozen birds, and they each have distinct voices and calls. You can always tell who just finished laying by which egg song you’re hearing, until the others join in and it’s a cacophony.
We’ve got one anericauna who likes to “scream” for her egg song. Awful sound, very close to a hen screaming because a bobcat got her.
I don’t think that they think that. All humans have different voices, yet it’s pretty easy to tell what mood someone is in, even if you don’t know the languages. Why would you assume that chickens express these so differently that it can’t be interpreted automatically, when the same is possible for us with much higher ranges of possible expressions?
It wouldn’t have to be a universal voice, just something consistent in the noise. Neat thing with lots of AI training is we don’t have to necessarily even have a theory of what the AI is picking out. It just has to be in the training data and our hints during training have to be appropriate.
Many many years ago I had a small contract with an animal research group making some of the gear for doing this with dogs and cats. My understanding is the project didn’t get great results. They basically could make predictions that were slightly worse than a dog trainer could.
Honestly, any decent dog owner can read the signs that their dog makes in terms of basic communication. We co-evolved after all. I don’t know how much a dog translator would be especially useful.
Adrian (the PI) is such an unusual, mad genius kind of guy. Peer reviewed or not, it would be cool if he and others managed to help us understand animals better.
If they're not getting consent from those chickens to use their clucks as training data they're engaging in animal abuse. Stealing their voices, how cruel.
They’re violating the chickens intellectual property by training the model on the chickens sounds without legally binding consent from the chickens to train on their data.
Wouldn’t be surprised if the chickens banded together and sued for copyright infringement.
Sounds like utter bullshit to me, but we’ll see what the peer review process comes out with.
Nobody should be reporting on studies that haven’t been peer reviewed. It’s often sensationalist nonsense, much like drugs that haven’t gone through clinical trials yet.
Add comment