Sure! That’s a large part of being a scientist. I’m running one at the moment.
I have two different AIs which are competing against each other to see which is better at recognizing people and places in historical documents. I already know who all those people and places are, because some people already figured it out by hand. We’ll see how close the AIs get to the humans’ answer. Whichever AI is closest the humans is the better one for this kind of annotation problem. This is a simple evaluation experiment.
Comments