See that em-dash after “years”? That tweet was written by an LLM. No wonder it makes no actual sense, an AI model has no idea that it’s ridiculous to berate a weather report, it’s just following instructions to reply with “AMERICA GREAT” to anything it thinks is “unpatriotic”.
thank god we still have that little tell. They’ll remove it and at some point we won’t fucking know what’s AI except when a moron suddenly says something coherent.
That’s kinda overly anthropomorphising. They are a statistical model that uses the number of occurrences of tokens in association with each other to predict the most likely next token. You can do that with pretty much any training data and any text. The current very large models will have almost any kind of text in their training data, including unhinged manifestos, racist tweets, etc.
Since models are stateless, getting them to produce that kind of output doesn’t change or worsen them in any way.
It’s not that simple. You can adjust outcomes by weighting. You can adjust outcomes in a myriad ways. Yeah, sure, in a trivial sense it’s just statistics, but that’s like saying that computer communication is just a bunch of ones and zeros. It misses a lot of the point.
It really gets way more complex and most certainly can be adapted and tuned in various ways.
See that em-dash after “years”? That tweet was written by an LLM. No wonder it makes no actual sense, an AI model has no idea that it’s ridiculous to berate a weather report, it’s just following instructions to reply with “AMERICA GREAT” to anything it thinks is “unpatriotic”.
thank god we still have that little tell. They’ll remove it and at some point we won’t fucking know what’s AI except when a moron suddenly says something coherent.
Oh yeah, the emoji too. Fffuuuck
It unfortunately does make sense, it’s just… VERY polarizing and inflammatory and sensitive and defensive
Scary how LLMs are taught to output this kind of offensive text
LLMs are trained and owned by rich billionaires, of course they protect the fascist class
That’s kinda overly anthropomorphising. They are a statistical model that uses the number of occurrences of tokens in association with each other to predict the most likely next token. You can do that with pretty much any training data and any text. The current very large models will have almost any kind of text in their training data, including unhinged manifestos, racist tweets, etc.
Since models are stateless, getting them to produce that kind of output doesn’t change or worsen them in any way.
It’s not that simple. You can adjust outcomes by weighting. You can adjust outcomes in a myriad ways. Yeah, sure, in a trivial sense it’s just statistics, but that’s like saying that computer communication is just a bunch of ones and zeros. It misses a lot of the point.
It really gets way more complex and most certainly can be adapted and tuned in various ways.
This LLM took some X patriot’s job god damn it
Russian troll factories in shambles.
Jesus