By being just a little weird, but in a pattern, and not to you. A pattern like: every n tokens raise the temperature (randomness) for one token. Then to dectect AI, you tokenize and calculate how expected each next token is. Then you try fitting the pattern to that.
Is it webscrapers?
deleted by creator
Supposedly the people paid to rate responses are already using AI
LLM is fine to say, not more awkward than element. And it is AI.
And I wish we’d call phones computers. It’s been very profitable for corps for people to not realize that they are.
Don’t forget the AI bubble is also a ram and SSD bubble.
Not in a “Darwinian moment”, like the CEO says. Then healthy social structures can be decisive.
Survival of the fittest is about species, not individuals. He’s saying AI is a threat to the human race.
Except it is a fairly good artifice of grass.
Good for what? Looking at? It’s not gonna satisfy a goat.
We only care for the look (“the output”), and we don’t expect more of it, or sell it for more than it is. That’s why it’s not a controversial term for astroturf. It wasn’t controversial for AI either until very recently. In 60 years of AI nobody has split hairs over output of intelligence. It’s justified but weirdly misdirected anger.
AI is a perfectly cromulent word for the thing.
Do you question the grass part of artificial grass? It’s obviously not grass.
Imho nothing wrong with AI use by professionals, as long as it’s verified. That obviously wasn’t the case.
BBQ your own scrolls
All AI is actual AI. It doesn’t need to be real intelligence to be artificial, should be obvious. Are you telling me artificial grass shouldn’t be called that because a goat can’t eat it?
Inference and training are separate in every ML architecture, what are you on about? And yes LLMs are ML, by definition, no need to argue.
Nah, this story has been developing since before the LLM boom. “AI” has never been rare in science headlines like this.
Our company decided to build our own ai translation system because the human translators we’ve been hiring started using AI… Quality dropped immensely, trust is lost. CEOs don’t feel like shopping around. So sad.
I was lucky I was forced to upgrade to 32gb right before the bubble, because my new job uses Jira with too many plugins.
No difference. Distillation is a valid and useful way of generating data to improve or make new models. It’s still just example data to be trained on. Anthropic is doing the same with their own models, and inadvertently every other model through web scraping.
The legal difference is that this data is uncopyrightable. At most it’s a TOS breach, nothing major.
A24’s success was always its marketing scheme: make all the movies, advertise the good ones. Once people started having expectations about the upcoming releases it was doomed.
Lower temperature is not necessarily better quality. At 0 it will get very repetitive, it’s for classification tasks, not for prose.
I suppose the optimal temperature depends on the model and the task, and I don’t think it’s very sensitive. Varying temp might even give better results, who knows.