It is worth considering how many training images an AI has been exposed to. Some experts have suggested that image classifiers need 1,000 images of each species before they mostly get it right, others have suggested100,000 are needed. (I expect it depends a lot on the individual cases). I guess Carbon has probably had only a few score of pigs so far, and almost no cattle. Until it has had sufficient of BOTH pigs and cattle (or any other pair of species) it will classify images as the nearest thing it has been trained on. If it had previously seen only wombats, I suppose Carbon would classify these as wombats, as well as classifying Christmas Beetles as wombats. Has Carbon been told that there is an upper and lower size limit to wombats, or that local wombats are not usually black and never shiny? No, because Carbon does not learn like we do. If you expect an AI to be able to identify a thing, there is a simple requirement: you need to feed it a lot of the thing.
In defence of carbon A1, it based id on initial photo, a distant one from my phone. I loaded photo from my camera shortly after. I have location data on phone only.
1,893,031 sightings of 21,044 species in 9,272 locations from 12,889 contributors
CCA 3.0 | privacy
We acknowledge the Traditional Owners of this land and acknowledge their continuing connection to their culture. We pay our respects to their Elders past and present.