Ms Rosita Apaza Machaca:

On the first part of the question, the failure of AI to recognise, say, pedestrians or certain situations is a lack of input in the dataset that is already there. We have to teach AI to recognise things better. That means working at the edges of the bell curve, getting input from more people with disabilities whether that is through a survey or somehow engaging the public in inputting data about what difficulties they experience and how they interact with AI so that the feedback can be more complete. I am well aware, for example, that when I am navigating, AI will not recognise certain things. For instance, if I am looking for a post box I will drop a pin to mark it but the information the AI is provided has to be accurate and it has to be doubled-checked because AI can give spurious information that is completely inaccurate and has no practical benefit. When it gives feedback it can be completely irrelevant to what you asked it sometimes.