Interesting, after Connie was talking up the AI written stories on the finance section of CNet on the show, today she made a statement that the AI had made fundamental mistakes in the text it had written - and the editors missed the misinformation as well.
For example the AI said that 3 percent compound interest in $10,000 would be $10,300 after 1 year. In fact, the total in the account would be $10,300, but the compound interest would be just $300, a big difference (and I nearly missed it at first glance, because the total is correct, but it isn’t talking about the total, just the interest). There were a few other mistakes that have been noted and all the content is being reviewed and affected articles re-written.
I know she said that it only wrote some filler information, so they could concentrate on the “real” stories, but it is a bit of a letdown all the same. I feel it is a shame on the one hand, but not surprising on the other.
More worrying is the news from Check Point Security, that Russian hackers have been experimenting with ChatGPT to see if it can write polymorphic malware. Again, not really surprising, considering it has already been used to write code for other things, including the LastPass vault analyser that Steve talked about on his show last week.
I would say, I’m not surprised and it was inevitable, in these early days, but it is a shame all the same.