A wonderful illustration of (one of the reasons) I don’t like AI…
Google’s AI, when I ask it a question:
Google’s AI, when I ask it a question but only so very subtly hint that I’d like a particular answer to be true:
1 Like
At least you are happy now. Unless you’re an Ulysses user with an Apple watch.
LLMs have no concept of “truth” and therefore are incapable of “lying,” which implies deliberate intent.
They are, however, sensitive to subtle indications of what response is desired by the human questioner. Because they have no concept of “truth,” they have no ability to resolve conflicts between the human’s preferences and objective reality.
Edit: It’s not the phrasing of the question that makes the difference. Here’s what I got (no quotation marks):
Yes, Ulysses does have a companion Apple Watch app. It’s designed to work with the main Ulysses app on iPhone and iPad, allowing you to view and potentially manage writing projects from your wrist. The Apple Watch app is part of the same bundle as the iOS app, meaning you download one app for both platforms.
1 Like
I gave both your prompts to ChatGPT 4o and it said nope, no watch app both times.
Such joyful inconsistencies.
1 Like