Raw LLM Responses

There's also the commercial viability problem. The money predicted by analysts has not materialised because there still isn't a good use case for LLMs in most commercial businesses. The few businesses that have tried implementing LLMs into chatbots have had high profile problems. Most organisations don't want to pay the amount Microsoft wants to charge for Co-pilot. There are interesting applications, their recent demo around email summarising for example does have a lot of potential use, but that's not worth $30/month/user for most organisations. This idea that LLMs are going to replace web browsers is just utterly insane though. And I can see them getting a lot worse once the flood of absolute AI garbage that's spreading throughout the sources for the LLMs gets input to models, their advancements will stop. Or the courts will find against them in the copyright issue and they'll be forced to rip up the models and only train on a much smaller dataset putting the industry significantly further back.