Here are the draft documents for discussion at tables:
Jon
DRAFT Algorithm Threshold Assessment - 10 July 2023.docx (58.0 KB)
DRAFT Algorithm Impact Assessment - 10 July 2023.docx (84.9 KB)
DRAFT AIA User Guide - 10 July 2023.docx (400.8 KB)
Here are the draft documents for discussion at tables:
Jon
DRAFT Algorithm Threshold Assessment - 10 July 2023.docx (58.0 KB)
DRAFT Algorithm Impact Assessment - 10 July 2023.docx (84.9 KB)
DRAFT AIA User Guide - 10 July 2023.docx (400.8 KB)
Here is the link to Robynâs paper:
Regarding the NAIAEAG advice on Generative AI:
How can we know if solutions that we use save the data into LLMs? For example there are various voice to text dictation software solutions out there, some used by Te Whatu Ora already. At first glance these may not appear generative, but itâs possible they could use generative type engines to help predict what the clinician was trying to say if the words couldnât clearly be heard. And they may save the data entered into them to learn from it.
It would be hard for the average clinician user of these systems to know what type of AI tool drives the system they use and whether it complies with the Te Whatu Ora advice or not.
Should this be part of our clinical safety assessment of any new clinical digital systems (if we had a NZ version of the UK DCB0129 standard
How do we avoid not missing the existing clearly defined AI tools and how can we use the existence of these to remodel our existing landscape quickly in a novel and competent way beyond just the AI capability?
e.g. Cardiac sonography AI capability: Opportunity for improving access / training / remote supervision / distance oversight BUT we have a fragmented landscape of poorly integrated systems that mostly do not have functionally competent linked cloud platforms.
WHO in our new world will see and asses the need, make the strategic decisions for NZ then ensure enactment?
Has NZ/HNZ joined any international collaborations on AI in Health or are we working with any international jurisdictions on this area?
Has an international best practice AI strategy for health emerged overseas that you have seen?
Will there be a need for regulation of A.I. tools in the future?
Has there been consideration for Enterprise grade Gen-AI solutions that donât ingest your interactions? Bing Chat Enterprise for example doesnât store any data, Microsoft have no eyes on it, is encrypted in transit, and does a read only from GPT4. âNothing leaves your tenancyâ is their guarantee.
So nothing being used as âtraining dataâ?