I generally avoid the AI interactions as my estimation is that the AI-hype makes AI into a panacea which it is not. The proposed tool, however, seems like a plausible use of AI in its current state, so I decided to give it a try.
I selected a question I've answered which could benefit from some editing to make it less awkward to read and increase the clarity of the problem.
The results were, to be generous, abysmal.
Image may be NSFW.
Clik here to view.
The major issues include:
- The section I've circled in the screenshot contains information which was critical to my ability to grasp the intended inputs and results. Completely removed in the edit.
- The single line of code, the OP's current attempt, was mangled to make it invalid Excel rather than not working Excel.
- One of the key points in the question was that the OP was attempting to "derive" information from an ID number, while the edit makes it look like the OP merely wants to "separate" segments of the number.
- Once I arrived at the question, the OP was mostly unresponsive to clarification questions. After reading the question, several times, I was able to piece together what I knew with what they said and form a "problem statement" which I could then find a solution for. Much of the data needed to reach that understanding is stripped from the question by the edit. The paragraph preceding my circled section is one such piece of information.
While the question could have been easier to grasp with some edits or refinements, it was at least an answerable question. The suggested edit results in a question which lacks the data needed to understand the OP's problem, or even find a solution, let alone validate the potential solution against the desired results.
As someone pointed out to me, there's a good probability that the question and answer, based on age, were included, verbatim, in the training data for the AI used. To me that suggests the results ought to have been even better than if the question had been a novel one, as the questions it will be deployed against should be.
My final conclusion is that, while the proposed use is a plausible application of AI, it's not yet within the domain of AI's practical usage yet.
Abandon the process now, saving the idea, and the process, for later application once it is within the domain of AI's practical usage.
It is a good idea with bad timing.
