I believe the term “GIGO” has it’s origins in software developed or operating with erroneous data input (garbage) that resulting erroneous output (garbage)—AI software will be different since the garbage being input will be deliberate to generate the desired garbage being output.
Bingo !
AI models are only as good as their "training", which consists of giving it large numbers of inputs and comparing them to the DESIRED outputs, adjusting itself with each iteration.
I personally prefer the term "Machine Learning" which is much more descriptive and technically accurate.