One of biggest challenges for Plantz is providing our users with the assurance that our AI Agent's responses are accurate and can be trusted.
With all the debate about AI in healthcare, Plantz know one of biggest challenges is providing our users with assurance the responses are accurate and can be trusted. With this aim, we’re publishing the result of our latest evaluations. To do this, we generate 50 questions about medical cannabis, together with the Agent’s response.
The response can then be evaluated by comparing it to the ideal answer. By doing this consistently over a period of time we can see how accurate the Agent’s response is to a given question.
We can see the Agent responded to the question, ‘Is smoking the only way to use medical cannabis?’ with an answer we’re not completely happy with. However, the reason is nuanced – medical cannabis is illegal if smoked. However, it’s widely acknowledged this consumption method is still popular with patients. Therefore, we have to work on training the Agent to realise smoking is not to be encouraged but recognise patients still choose to do it. We’ll keep tweaking until we’re satisfied with the Agent’s response.
However, the other answers look accurate and we’re comfortable with the majority of responses. If you see somethi8ng you feel is inaccurate or could be improved, please let us know.
Group Forums
Discuss these articles by joining a Group
Latest Posts
Plantz is owned by Million Media Ltd, a registered UK company number: 15476153