The big detractor for me is its heavy curation towards predetermined outputs that are constantly being tuned by a small army, essentially making it a Mechanical

The big detractor for me is its heavy curation towards predetermined outputs that are constantly being tuned by a small army, essentially making it a Mechanical Turk paired with a nifty model. They can spin it as "trying to improve its usability", but it looks more like "trying to mask its unusability" to me. It's the world's most heavily invested-in parlor trick, and anybody who drops it into a situation of higher stakes will find that out quickly. The GPT-4 paper even warns about this: Most importantly, it still is not fully reliable (it “hallucinates” facts and makes reasoning errors). Great care should be taken when using language model outputs, particularly in high-stakes contexts, with the exact protocol (such as human review, grounding with additional context, or avoiding high-stakes uses altogether) matching the needs of specific applications. Whenever they create a model than can achieve the same outputs without the heavy curation and deliberate bias on certain subjects (like condescending people for asking where to find cigarettes), then I'll be impressed.