Upgrade your account to a Community Supporter account and remove most of the site ads.

Reply to thread

Message: <blockquote data-quote="Cergorach" data-source="post: 9862406" data-attributes="member: 725">Imagine how much the 'nailgun' does for wooden house building and renovation, if you give that same tool to the average person, they will probably nail themselves to their car with it...Just because you don't know how to use a tool or know how to choose the right tool, doesn't mean the tool is useless. Not all LLMs are created equal for the same tasks, often different providers have different versions available. And just like with every other product, people sell it telling lies (or do you really think dish-soap X will be so much better then dish-soap Y, like they show in the commercials)...No, but people are different in their unreliableness. People either lie or make mistakes, in most cases people leave hints in their lying to a greater or lesser extend. And people make mistakes a bit more predictably then LLMs. You know from experience that people are weaker in certain areas, and sometimes there are also additional identifiers when people make a mistake.The problem with AI (LLM) is that they do not lie and the do not make mistakes, the answer might be faulty, but that isn't a lie or mistake. They just follow their design, making a correct answer virtually undistinguishable from a faulty one, unless you know the material very well and are paying attention.The further issue with that is, that humans will be attended to their mistake and learn from it, the same with a lie. And if they do not learn from that behaviour, they will be replaced. The same isn't true of the current AI (LLM) products, they are the same for not just the same organization, they are the same for the whole world that uses the same supplier. They do not learn from the faults or the lies, nor will they be as easily be removed as humans.People aren't reliable either, that's why in many processes there is the four eyes principle, where multiple people check each others work, and in certain cases quite a few more people look at the thing that needs to be checked.Sure people can check the AI (LLM) work, but that requires discipline, knowledge and experience. And as we've seen in the news, it happens that these checks haven't been done. Either due to folks being lazy/lacks or folks doing things they have no knowledge in in the first place. These things of course happen with purely human situations as well, but they tend to not make the news anymore unless they are extreme cases. Currently reporting on the successes and failures of AI (LLM) is 'sexy' for the media...Some really old people around here might remember how administration or finances worked before computers and Excel, that required a LOT more people at the office. Getting a loan at a bank wasn't dependent on your actual income, other debts, dependebility, etc. But how well the branch manager knew and liked you... And if you fell outside the 'norm' at that time or didn't go to the same church (or any church), you were SOL...AI (LLM) will fill a similar niche imho. It won't replace everyone, and it's just a tool amongst many (just like Excel). And sure, there will be folks that will use it for stuff it isn't supposed to be used for, just like they've been using Excel as a shadow database application (something it was never intended to be used for) without telling anyone in IT about it... AI (LLM) will get misused, just like every other human invention...</blockquote>

Verification