> working with [chatbots] feels like groping through a cave in the dark – a horrible game I call "PromptQuest" – while being told this is improving my productivity.
Large language models are a single kind of AI, and a particularly annoying kind when you are forced to use them for deterministic or fact seeking tasks
or did you read the article? you're probably an LLM. why am I here? fuck this website
I don't think the future of AI is with LLMs either. Not only LLMs anyway.
This seems to be someone who has no idea how to use LLMs yelling at clouds. Or maybe just someone pretending to have no idea because it makes for good cloud-yelling.
How to use LLMs? There’s no “how to use LLMs”, you just tell them what you want, and they give it to you. Or they give you sometimes, and sometimes they give you something else. Or they’re telling you they’re giving you exactly what you want, but don’t. Or they seem like they’re giving you what you want, only it’s got a secret mistake somewhere inside, that you need to dig through and search for. Maybe there’s no mistake after all.
Yes, this is clearly a new wonder-technology, and all criticisms of it are just old people, back on their cloud-yelling bullshit.
The way I see it, when LLMs work, they're almost magical. When they don't, oh well, it didn't take that long anyway, and I didn't have them until recently, so I can just do things the old boring way if the magic fails.
But for tools, we should have a clear up front list of capabilities and menu options. Photoshop and VScode give you menu after menu of options with explicit well defined behaviors because they are tools used to achieve a specific aim and not toys for open ended exploration.
An llm doesn’t give you a menu because the llm doesn’t even know what it’s capable of. And that’s why I think we can see such polarized responses - some people want an LLM that’s a supercharged version of a tool, others want a toy for exploration.
Generative AI is neither magic, nor does it really solve any problems. The illusion of productivity is all in your head.
For my uses, my rule is "long to research, but easy to verify". I only ask for things I can quickly determine if they're right or not, I just don't want to spend half an hour googling and sorting though the data.
For most of my queries there's an acceptable margin of error, which is generally unavoidable AI or not. Google isn't guaranteed to return everything you might want either.
Idiots like this seem to want a robot that does things for them instead of a raw tool that builds sometimes useful context, and the LLM peddlers are destroying their creations to oblige this insatiable contingent.
I actually agree with the article that non-determinism is why generative AI is the wrong tool in most cases.
In the past, the non-determinism came from the user's inconsistent grammar and the game's poor documentation of its rigid rules. Now the non-determinism comes 100% from the AI no matter what the user does. This is objectively worse!
There’s chat-vs-api; same model answers differently depending on input channel.
There’s also statistical. Once in a rare while, a response will be gibberish. Same prompt, same model, same input mode. 70% of the time, sane and similar answers. 0.01% of the time, gibberish. In-between, a sliding-scale — with a ‘cursed middle’ of answers that are mostly viable except for one poisoned thing that’s hard to auto-detect…