I think this article is more about the future of AI, not what AI currently is.
Yes, AI has a lot of flaws now. However, many of them will be minimised in the future. It’s developing rapidly.
]]>It didn’t find those cases in publicly available data, it straight up manufactured them on principles of ‘that sounds right’.
]]>ChatGPT is trained on publicly available Internet data until about 2021. The same technology used on a proprietary legal database (think WestLaw or Lexis) would likely not cause the same problems.
]]>