Many workers are faking knowledge of AI to make sure they aren’t left behind::There’s a need for more AI training, report finds
Many workers are faking knowledge of AI to make sure they aren’t left behind::There’s a need for more AI training, report finds
I read an article the other day where an airline was breaking about using AI to predict how many passengers will buy a meal in flight based on how many people had historically bought a meal in flight.
That’s… Literally just an average of how many people order a meal…
Ehhhhh there are much more sophisticated models than just an average. What a neural network could do is derive inferences based on a wide variety of inputs like time of day, country of origin, individual passenger characteristics, and so on.
Ultimately that application is just averaging over a smaller subset.
While admittedly I don’t know that scenario myself, it looks like several scenarios I’ve seen where we imagined some magic insight from AI over more limited statistics, but not one of those scenarios ever predicted better.
That’s not to say AI approaches are useless, but this sort of data when the dataset is pretty well organized and the required predictions are straightforward, then a pretty simple statistical analysis is plenty, and declaring “AI” for such a simple scenario just undermines AI credibility where it can do formerly infeasible things.
You can basically think of AI as a massively multiverse analysis that can go far beyond a directly applied model. So while yes, technically averages are involved, they’re applied in a way that makes it incredibly naive to call it “just averages”.
Edit: it is especially not “just an average of how many people order a meal” as you had said.
AI models are averages, except in the form of weights over a large set of matrices. However, calling them “just averaging” is grossly oversimplifying how they work.
Most “AI” is just outcomes from machine learning.
And what are you ? Some statistics wizard ??