You can always address common concerns, such as privacy, data security, and hallucinations.
Whenever I use AI apps, I always disable settings related to model learning on my data, as I prefer not to contribute to the model’s knowledge base. 😉 In the case of Atlassian apps, you can highlight their isolation from other users to reassure customers.
If the AI is based on a well-known model like GPT-4, transparency is important. However, if it’s a custom or newly trained model, additional privacy and data concerns arise:
- Will it be trained on my data?
- If not, how and when will it be updated? (e.g., with the recent news about GPT-4.5 being released)
- Does it continuously learn from interactions, or is it static?
Another key factor is temperature settings:
- If I ask a question today, will I get the same answer as someone else asking tomorrow?
- Is the AI acting like a "teammate," adapting to conversations, potentially providing "expected" answers rather than strictly factual ones?
-
For example, GPT-4 can be lazy! Even when provided with links to documentation, it may ignore them and prioritize its own knowledge base. It can even claim to have searched the web while still providing incorrect answers.
-
New concern: How was the app's reliability tested?
On the other hand:
- Will I actually use it? Will my team use it, or is it just for the CEO?
- How often will I use it? Daily, once a week, or only once a month?
- Does it truly save time?
This is part of what I consider when evaluating AI apps. Send me the link to the article when you’re done - I’m curious to see if you’ll address any of these concerns and what the answers will be! ^^
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.