A user asked on the official Lutris GitHub two weeks ago “is lutris slop now” and noted an increasing amount of “LLM generated commits”. To which the Lutris creator replied:
It’s only slop if you don’t know what you’re doing and/or are using low quality tools. But I have over 30 years of programming experience and use the best tool currently available. It was tremendously helpful in helping me catch up with everything I wasn’t able to do last year because of health issues / depression.
There are massive issues with AI tech, but those are caused by our current capitalist culture, not the tools themselves. In many ways, it couldn’t have been implemented in a worse way but it was AI that bought all the RAM, it was OpenAI. It was not AI that stole copyrighted content, it was Facebook. It wasn’t AI that laid off thousands of employees, it’s deluded executives who don’t understand that this tool is an augmentation, not a replacement for humans.
I’m not a big fan of having to pay a monthly sub to Anthropic, I don’t like depending on cloud services. But a few months ago (and I was pretty much at my lowest back then, barely able to do anything), I realized that this stuff was starting to do a competent job and was very valuable. And at least I’m not paying Google, Facebook, OpenAI or some company that cooperates with the US army.
Anyway, I was suspecting that this “issue” might come up so I’ve removed the Claude co-authorship from the commits a few days ago. So good luck figuring out what’s generated and what is not. Whether or not I use Claude is not going to change society, this requires changes at a deeper level, and we all know that nothing is going to improve with the current US administration.
I feel like there needs to be a dedicated post (and I don’t want to write it, but maybe I eventually will) that outlines what a model really is. It is not just a statistical text prediction machine unless you are being so loose with the definition of “statistical” that it doesn’t even mean anything anymore.
A decent example of a statistical text prediction machine is the middle word suggested by your phone when you’re using the keyboard. An LLM is not that.
In the most general terms, this kind of language model tokenizes a corpus of text based on a vocabulary (which is probably more than just the words in the dictionary), uses an embedding model to translate these tokens into a vector of semantic “meaning” which minimized loss in a bidirectional encoding (probably), that is then trained against a rubric for one or more topic area questions, retrained for instruction and explainability, retrained with reinforcement learning and human feedback to provide guardrails, and retrained again to make use of supplemental materials not part of the original training corpus (resource augmented generation), then distilled, then probably scaled and fine tuned against topic areas of choice (like coding or Korean or whatever) and maybe THEN made available to people to use. There are generally more parts to curriculum learning even than that but it’s a representative-ish start.
My point being that, yes, it would be nuts to pose ANY question to a predictor that says “with 84% probability, the word that is most likely follows ‘I really like’ is ‘gooning’ on reddit”, but even Grok is wildly more sophisticated than that and Grok is terrible.
Edit: And also I really like your take at the start of this thread: user error is a pretty huge problem in this space.
The training is sophisticated, but inference is unfortunately really a text prediction machine. Technically token prediction, but you get the idea.
For every single token/word. You input your system prompt, context, user input, then the output starts.
The
Feed the entire context back in and add the reply “The” at the end.
The capital
Feed everything in again with “The capital”
The capital of
Feed everything in again…
The capital of Austria
…
It literally works like that, which sounds crazy :)
The only control you as a user can have is the sampling, like temperature, top-k and so on. But that’s just to soften and randomize how deterministic the model is.
Edit: I should add that tool and subagent use makes this approach a bit more powerful nowadays. But it all boils down to text prediction again. Even the tools are described per text for what they are for.