CHS is very readable. He has free samples of his books on his website. Very logical arguments, good social criticism. But I wonder if AI model training can recognize and discard the AI slop that is cluttering the internet now. It can be maddening. A recent slop piece on a stock that I own was titled "[ticker] drops 90% ..." after using 5-year old data from the highest point of a bubble. How much more of this will we see from unsupervised AI slop publishing? Will automated trading algorithms processing news feeds be able to recognize irrelevance? Just one example...
CHS is very readable. He has free samples of his books on his website. Very logical arguments, good social criticism. But I wonder if AI model training can recognize and discard the AI slop that is cluttering the internet now. It can be maddening. A recent slop piece on a stock that I own was titled "[ticker] drops 90% ..." after using 5-year old data from the highest point of a bubble. How much more of this will we see from unsupervised AI slop publishing? Will automated trading algorithms processing news feeds be able to recognize irrelevance? Just one example...