How AI Is Changing Shopping: Product Recommendations | Oliver Edholm

Here's a summary of the great stuff that we cover in this show:

• If you want to get started out with AI, your traffic and the number of SKUs that you have must be at a reasonable level before AI makes sense for you. Several 100 SKUs and probably a couple of 1000 people at least visiting your website. The more the data, the better the AI tends to be the rule of thumb.
• Depict can work with more sparse data sets, which have less data. Product information is incredibly useful when recommending a product like going to a physical retail store and asking the salesperson for advice. Almost all existing recommender systems ignore that. Depict can apply incredibly smart image recognition algorithms, understanding subtle patterns, which was impossible a few years ago, but is now possible. When combined with the user data, it can generate really good results.
• There are a lot of buzzwords around AI and many people exploit that fact. The way to avoid gimmicks then is to become really good at asking yourself, what problems do I need to solve and does the solution actually solve this problem? And then don't care if it uses AI or not. Simple solutions tend to work surprisingly well, a lot of times.

• Superintelligence (book by Nick Bostrom)
• Klarna