This.
Fun fact: Back when LLMs were new and people didn’t realize the ethical nightmare that occurred to get the training data, I discovered that chat bots actually set off my social anxiety quite severely, to the point that I couldn’t use one, even if one were trained ethically.
A much more accessible resource for people like me would actually be static websites and tutorials. I don’t think chat-bot-as-instructor is a very universal solution, contrary to the topic’s proposal.
I have social anxiety that is occasionally, although not always, like a monster gripping my head. Chat bots don’t trigger it (although I’ve found them intensely boring), but the idea of extended chatting online does. Every time someone invites me to one of those running chat thingies, I feel the monster’s claws. So interesting how differently such things manifest.
I find it hard to take Chat bots seriously. From everything I’ve gleaned from the smart people, they aren’t good at much of anything and won’t get better anytime soon. Isn’t this another one of those fads that will just die out in a year or two when they don’t reach their potential?
There are definitely different levels of experience and results. Anyone that thinks this is a fad is mistaken. Overhyped, yes. Bad data training, definitely. Resource intensive, yes especially where energy consumption is concerned. The ROI compared to what’s been invested is unlikely to be recognized.
But if you want to build a react website with simple features, you no longer need to know React or node or anything. GenAI will walk you through everything.
GenAI is excellent at organizing your documents and summarizing them.
I agree the creative things in the world should remain human-sourced, but GenAI can be an excellent assistant.
I cannot imagine that the results of this experiment would be worth the resulting reputational damage to IFTF and its projects.
I agree, but man you just made it creepy.
I’ve used AI for help coding when I need to understand how some poorly documented framework API is intended to be used or to get ideas of possible approaches. But…it’s really only helpful if you already have a lot of contextual knowledge around the topic you’re examining, otherwise you’re just as likely to get suggestions that are plainly (or far worse: subtly) wrong.
One thing I think an LLM is good for is as a concierge for a walled-garden wiki or database.
Many users of self-support may not know what keywords will find the wiki they want - in my job there’s an article about the supporting documents a provider needs to submit for certain types of claims. The best way to get this article is to type ‘supporting documents’ but for a provider or agent who doesn’t know why the claim is refusing to submit likely doesn’t know why and might not know what keywords to search for. The AI might be able to decipher a question like “why can’t provider submit a partial claim” which through standard search will pull every article with the words “provider” and “claim” but a trained AI might understand through experience supporting documents are required and offer the right article, or just pull out “Make sure the provider is including documents X and Y when submitting…”
Google search’s generative AI is good at parsing a piece of data out of extensive results. “What are Millenials?” has a number of different varying opinionated responses, but AI is good at summing up:
Millennials are people born between 1981 and 1996, and are also known as Generation Y.