I am not so sure about the generative models thing. Everyone is losing their shit about ChatGTP et al, yet I am sitting here saying yeah neat trick but I canno

I am not so sure about the generative models thing. Everyone is losing their shit about ChatGTP et al, yet I am sitting here saying yeah neat trick but I cannot trust anything it generates since anything and everything can be 100% made up and totally inaccurate and outright false. It is neat technology, but ultimately it is like a dictionary where you have to double-check that it is telling you the correct spelling every single time you look up a word - totally useless in other words: why bother using the dictionary at all if you then need to go check that the dictionary is right using some other source of truth? Why not just go straight to that and skip the dictionary in the first place since it is so untrustworthy. Every single time I've asked ChatGTP or Bard a "simple" question that I know the answer to, there has always (always!) been something simply outright wrong. It is just not ready to be trusted right now, but I look forward to the day when hallucinations are solved. Until we can be sure that LLMs do not hallucinate, I struggle to see the value to everyday-people for LLMs beyond use cases where accuracy and facts are unimportant. So ignore google-killing search engines, ignore medical diagnosis, ignore programming, ignore legal advice, and instead embrace making up stories or poems about unicorns or whatever where facts, accuracy, and correctness simply don't matter. It will still have a big impact sure, but I don't think it will be as world-changing as people think once the novelty wears off and people realise that LLMs are generally not very reliable when it comes to giving accurate info, and are better suited to "creative" tasks where correctness/accuracy doesn't matter. E.g. with stable diffusion (remember that?) no one expected a perfectly anatomically accurate picture of a dog or whatever - it did a passable picture of a dog eating a burrito or whatever the prompt asked for, but no one confused it with reality or treated it as fact. This is where I feel that we are with LLMs right now - passable, but does not hold up to detailed scrutiny. But obviously yes the metaverse is obviously total bullshit and facebook have totally jumped-the-shark on this. They're currently deep in restructuriung to undo over-hiring during the pandemic - it would not surprise me at all to see them also re-pivot away from the metaverse too (couching it as some sort of "focusing on efficiency" or whatever, i.e. continue working on FB, instagram and whatsapp core features)