It doesn’t take special training with a butcher to look at these four chicken breasts and tell which one you’d want to eat.
At the same time as the meat quality has suffered, activists have pushed the industry to move birds to more open, social environments, ones where they aren’t caged up all the time. In those new settings, new behavioral problems have emerged. Some birds have turned out to be very aggressive, engaging in vicious pecking of their fellow fowl. “If they are in big groups and don’t know each other, they’ll peck each other. Sometimes they’ll peck each other until they rip out all each other’s feathers,” Baes said. “It can progress even to cannibalism.”
That’s the future that turkey breeders want to avoid, while also solving the problems they already have. “How do we make sure they can walk, that their bone integrity is good, that they don’t kill each other? Do we just breed for a very docile bird?” Baes asked. “Do we breed for shorter beaks, so when they do peck each other, they don’t do as much damage?”
Doing any of that requires data. Baes gave me an example of how much harder it is to gather this kind of information than the traditional traits that breeders have optimized for. Let’s say a turkey farmer wants to measure how well her turkeys walk. The farmer has to individually grab the turkeys and have them walk. Seems easy enough, but let’s say the flock is 6,000 birds. That becomes a massive data-collection task. If the farmer wants to measure for, say, aggressiveness, she could deploy some kind of machine-learning system on video of a flock, looking for the aggressive birds that peck others a lot. But living creatures are complicated.
“We could have a heyday with machine learning and algorithms, but at the same time, there is the biological basis behind it,” Baes said. “Looking at video data, an aggressive peck at the head of another bird is going to be very different from a gentle peck at a piece of sawdust that the bird sees on the ground.”
Nonetheless, the industry is trying all kinds of sensors, automated video analysis, and other technologies to try to find good measurements for the traits it is trying to improve.
In one experiment, Baes’s team placed a little force plate under a feeding area. A bird walks in, gets an RFID tag scanned, and then the scientists measure how hard it pecks at the feed they’ve given it. They see a very clear difference in pecking intensity among different birds. “It is so cool,” Baes told me. “We are entering a new realm of data collection and of animal breeding and of all of these technologies coming together, hopefully being put to use for a good cause, like breeding more timid birds that are not gonna beat the crap out of each other.”
The breeders know how to make animals change. They build associations between genotypes and the observed phenotypes. Then they start to predict how cross-breeding different lines of birds will change the animals. Once they’ve got a measurement to underpin their systems, they will force the evolution of the livestock. It takes years for the genetic enhancements to filter from what are known as “pedigree” lines down to the commercial stock, but it happens.