[Wednesday 3/17: 3 miles; Saturday: 6.1 miles; today: 3 miles]
In my current job I am sort of on the edge of some machine learning work. Machine learning is a thing where you build a pattern recognition system, then you throw samples at it and you tell it the right answer for each sample and it slowly “learns” the fine differences between the positive and negative samples.
One of the difficulties in this is having enough examples, both positive and negative, which means just a ton. And they have to be pre-scored by some other process. Frequently it means hiring a bunch of people to look at the samples and score each one, which is a boring and probably not well paid job.
If you’ve been on the internet you’ve seen this kind of thing. You go to a website and you have to prove you aren’t a robot and to do that they show you a scene and you have to pick the parts of the scene that have traffic lights or cars or pedestrians. You’re scoring the picture that someone else has previously scored… so they let you in.
So you can imagine doing that all day with piles of pictures and no one to tell you the “right” answer. Maybe I would rather flip burgers.
I saw this notice posted at the Cabella’s in Prairie du Chien yesterday. We went up there to look at their fishing supplies. This probably isn’t a machine learning project, but it could be the collection of a sample set they could use for such a thing. I have no idea what they will use it for. The straightforward use is to help them count the various forms of critters that show up in various places. But if they save the pictures and save the choices, they could teach an ML system to make similar choices.
I wanted to add that there are other similar sites. One that has been around a long time is the Chicago Wildlife Watch.