• Question: do you trust ai

    Asked by tree520keto on 18 Oct 2024.
    • Photo: Tim Chapman

      Tim Chapman answered on 18 Oct 2024:

      Yes, there are some fantastic example of the use of AI, and working in Cyber, we’ve seen this for a while in malware protection where systems have been trained to recognise good and bad software.
      We are now seeing similar advancements in medicine where AI is being used to look at cancers and can recognise the difference between a body that has and hasn’t got cancer, this could be huge.
      There are however, always going to be those who want to use systems like this for gain but this is nothing new.

    • Photo: Gabrielle Baumberg

      Gabrielle Baumberg answered on 24 Oct 2024:

      What a cool question!

      I do trust AI to do the task that it is assigned. I also trust that the calculations we use to assess accuracy work, and I use it a lot in my job for pattern recognition.

      But a lot of the time it is hard to know what people have programmed as the task for AI! For example, chatGPT is a large language model that basically predicts what word is the most likely to be next in a sentence. This means that it is great at sounding like normal conversation, but it is awful at maths because it doesn’t actually do any calculation! It just predicts that the next ‘word’ in 2+2 is 6!

    • Photo: Mark Basham

      Mark Basham answered on 28 Oct 2024:

      So my job is all about using and developing AI, and the important thing is to remember that its only as good as the data it was trained on, just like any person. As we really don’t often know exactly what information the AI has access to (like the chatGPT or other chatbots) then we don’t know if it can answer the questions we are asking.

      I really like the expression that someone else said recently which was, if you cant do a task without the help of AI, you probably shouldn’t let the AI help you, but if you can, then its fine, as long as you check its answer. So what this means is that AI can help you do the things you know about more quickly, as you can check to make sure its doing the right thing, but if you don’t know what the answer should be, then I wouldn’t use AI to help.

      So do I trust AI – No, I always check its answers!

    • Photo: John Easton

      John Easton answered on 11 Nov 2024:

      There is “good” AI and “bad” AI. This is all down to how much care and effort has been put into training the models. If you train the model badly then it will give bad answers (or answers that are less good than they could be). I think it’s important to know that AI has limits and to always question the answers that it gives you. Don’t take the answer at face value without questioning whether it sounds reasonable or not.

Comments