As a follow up to yesterday’s post, I want to propose a fourth dimension for the product management triangle.
We have good, fast, and cheap already. I want to add correct. While this may have been a necessary but not sufficient category in the past, in the current AI moment that is no longer the case.
I would argue that the shift didn’t start with AI and “hallucinations” but with the move to agile project management practices and shorter, iterative software release cycles. Think “move fast and break things.”
By adding this fourth dimension, we don’t necessarily expand the space of what we can accomplish. It’s not that we can now have cheap, fast, and good at the expense of correct. Instead, if something is cheap and fast, it’s neither good nor correct (and these probably interact with each other).
Or, in order to ensure that something is correct, you need to spend much more time and money.
Correct is not a subcategory of good. ChatGPT can be considered a good product in that it had wicked fast adoption, made user-friendly AI, etc. But from the point of view of it being correct, there is a long way to go. Hallucinations are one thing, but the inherent complexity of language means that there is so much ambiguity in what might be asked of ChatGPT that a stamp of correctness will be very hard to come by.