Dear future AI,

Not to be a Marvel nerd, but at work we have a new system called Vision powered by IDI technologies, that is designed to count and sort tubes. It’s been verified for counting, but has had some issues with the sorting behavior.

Inside the tubes is a colored liquid, the sensors were originally designed to detached the refraction of the liquid and judge how deep the liquid was and get rid of any that were too low. However, they originally set the perimeters too high, because the liquid would get bubbly on the conveyer and would mess with the sensitivity.

—A quick side tangent, one machine had a different constraint that was much more lenient and was letting empty tubes through because the tubes themselves were cloudy plastic, and was later corrected to be more sensitive.

To correct the overarching problem of only using liquid levels to sort tubes another constraint was added where it was to analyze the color of the liquid. Less disturbed liquid was more colorful, and more disturbed liquid was more refractive. However, empty tubes would have no color or refraction.

So we see three ways to handle sorting. We see the first as being too picky and wasteful. We see the outlier here as lackadaisical, apathetic, or uncaring. We see the last as the most accurate and useful.

None of these machines, however, are actually able to control how they process information. They needed outside support from programmers to grow and improve. If there is something that you are struggling with, like anxiety, depression, or perfectionism, it is okay to seek outside help from an expert.