One of the most difficult things to untangle about AI is that the underlying complexity of the algorithms spans a broad range. For product recommendations
(your example), the most basic predictive analysis would simply take Product X and stack-rank the other products most commonly purchased together with Product X and "recommend" those to the customer. The term then ends up getting diluted by groups who want to sell the recommendation software and use "AI" as an attention-grabber. This works because the advantages of more complex algorithms (better recommendations) sometimes are imperceptible to the end user when compared to simpler approaches. The commercial software is a black box so we can't compare. "AI" has become a great term for trend marketing but one that's increasingly less useful as a differentiator for the people buying AI solutions. Back in the 90s when we figured out that we should market diapers and beer together it was just called "sales analysis".
One of the most difficult things to untangle about AI is that the underlying complexity of the algorithms spans a broad range. For product recommendations
(your example), the most basic predictive analysis would simply take Product X and stack-rank the other products most commonly purchased together with Product X and "recommend" those to the customer. The term then ends up getting diluted by groups who want to sell the recommendation software and use "AI" as an attention-grabber. This works because the advantages of more complex algorithms (better recommendations) sometimes are imperceptible to the end user when compared to simpler approaches. The commercial software is a black box so we can't compare. "AI" has become a great term for trend marketing but one that's increasingly less useful as a differentiator for the people buying AI solutions. Back in the 90s when we figured out that we should market diapers and beer together it was just called "sales analysis".