• Question: How do AI algorithms handle uncertainty and ambiguity in data?

    Asked by date519sank on 20 Feb 2025.
    • Photo: Aurelia Brzezowska

      Aurelia Brzezowska answered on 20 Feb 2025:

      When AI is given data, most of the time it is “cleaned.” This is a process of removing ambiguous, incorrect or empty data as to not confuse the AI. Luckily AI is built on ambiguity. When it sees a picture of a puppy, it will compare it to all the data is has been given and give a percentage estimation of what it thinks it is. You may see its output as it thinking its 25% dog, 53% cat and 22% painting. The better the data quality, the more correct it’s answer will be. AI is only as good as what it is fed!

    • Photo: Neil Barnby

      Neil Barnby answered on 25 Mar 2025:

      This depends on the type of model you are using for instance Bayesian Networks use probabilities, so it doesn’t always give one answer it may say there is a 65% chance it is this and a 35% chance it is that. Some models use fuzzy logic which is rather than yous or not it has degrees of truth so with a blurry image and a question of is it a cat or a dog it would say it is 70% cat and 30% dog. Some models require more examples, so a larger data set or even human intervention.
      The thing to remember is AI doesn’t always have perfect answers.

Comments