AI ain't that smart after all

The predictive model is making conclusions based on data which was not presented (noise under load) and basically doesn't really focus on the bugged performance anomaly where the more expensive card is slower than the cheapest option.

To write "fluff" it does pretty good job and with some human reviewing, can certainly allow tech writers spend more time to do what they do best:

test hardware

By "massaging"' the chatgpt prompts and end results, you can create pretty well fleshed out articles which are not boring per se, even generating charts left and right.

However for well thought out opinion pieces and conclusive thoughts there is real value of having the actual tech writer intervene and put his own analysis in there, the predictive AI doesn't actually learn/understand what it is writing after all.

It will be interesting where this road leads us, ignoring it is something we definitely don't want to do. Combined with deepfake videos, text to speech and these AI modeling engines, it opens up a new world of creative tinkering. As with everything tech related, it can be used for bad, or good.