
A computer deciding if a cat or a bird was on camera used to be called computer vision.
Giving the computer this perception was done with machine learning, and perhaps to achieve this, some big data algorithms got used. Artificial Intelligence used to be reserved for a computer model that could pass the Turing Test.
It is interesting now that pre-trained transformer models can pass the Turing Test; everything is “AI”.Pre-trained models have a brutal simplicity to them. Train model ‘B’ with a higher number of parameters and a larger data set than model ‘A’, and model ‘B’ will be better than model ‘A’. Training model B will simply require more hardware capacity, more time, or both of those things.For the first time in generations, the capability of software is dependent on the quantity of hardware available to the people creating the software.