“Machine-learning” has enormous positive implications for livestock production, said an American swine expert.
But machine-learning, which is a form of artificial intelligence in which computerized systems learn from experience rather than simply following commands, relies completely upon heavy human input and interaction.
“The computer doesn’t know until we as humans put our expertise down,” said Tom Stein of Maximus Systems.
“The foundation of machine-learning is a lot of data. Lots and lots of data. But it has to be the right kind.”
Read Also
Beef check-off collection system aligns across the country
A single and aligned check-off collection system based on where producers live makes the system equal said Chad Ross, Saskatchewan Cattle Association chair.
Stein told farmers at the Manitoba Swine Seminar in February that machine-learning was already developing systems that solved real and challenging problems in hog production.
He has advised some young developers in creating a system that can identify the sound of a piglet being crushed by a sow, and differentiate that sound from all the other squeals, grunts and screeches common in hog barns.
It can then trigger a vibration in the sow pen, which should make her rise off the piglet, or if that fails it can send an electrical stimulation, which has about 99 percent success in briefly getting the sow to rise and free the piglet.
Machine-learning can also be used to identify the signs and behaviours of illness, which is something that demands a lot of monitoring by humans today and can have major financial consequences for barn operators.
Many areas of hog production can benefit from machine-learning, Stein said, but moving into the area shouldn’t be done willy-nilly.
Integration of multiple systems needs to be done carefully to both make the systems compatible, and allow barn workers and managers to handle the systems management without getting overwhelmed.
Stein said big acreage farming is ahead of livestock production in machine-learning.
