neural networks do their best to answer your nonsensical questions in earnest and they're adorable @OCRbot
@rugk these are just screenshots from talktotransformer.com because i can't be arsed to figure out how
@strnom yeah, wow, even philosophical/political takes are great here. Only a little contradictory 😜 (just when you claim the one thing is a big problem in one sentence and decline this thing exists in the next lol)
@rugk yeah, it can kind of sort of keep track of the overall topic it's talking about, and sometimes even of what its own take on it is, but it tends to lose track of "muddier" details such as whether it's speaking in a formal tone or not. it can, say, slide from academic to trumpspeech and back over the course of a few sentences
@strnom and it gives you creepy links sometimes
@rugk oh yeah! it likes to hallucinate links to things. for example, if it's talking about some facebook group, it will often add a link to
@strnom 🤣 Yay. But only if you know _nothing_ about animal welfare. If you know something about it, you may have heard something 😜
@rugk @simon the Leave.EU campaign created thousands of variations of a few memes, most of them catered to a single person at a time, who happened to live in a relevant constituency and were deemed influentiable enough on the brexit issue. they did this by hand, by looking at their facebook posts and likes and making educated guesses on the arguments that would sway them.
now imagine what they could do if they could straight up feed your facebook posts to a computer program, and it returned an incredibly persuasive, customised argument/meme/whatever on why you specifically should vote leave
Trying to detect an AI generated text by another AI is basically GAN, so this method is guaranteed to make the fake news generator even better over time, in an infinite cycle. Also, this is exactly how an AI might become sentient in a very short time. Look what a GAN can do: https://invidio.us/watch?v=kopoLzvh5jY (also from OpenAI)
(sry, for German context)
@strnom if you give it good hints, it can come up with funny stories… 😃
Yeah, who needs creativity when we have AI. 👍 😆
@rugk we can have both! we can have AI-powered creativity. this is genuinely why i'm so into this shit
@strnom ohh part 2 of this story features snape…
@strnom nearly classical
@strnom I like that story. It subtly touches so many issues… 😄 (and get's a little conspiracy 😆)
@rugk "november 9" i mean you can only be that transparent with the 9/11 allegory
@strnom *The one* server!
@strnom okay, so I'll stop here. But it's indeed funny. 😆
@rugk still a better story than fantastic beasts 2
@strnom okay okay, only this one!
The social network of the future: No ads, no corporate surveillance, ethical design, and decentralization! Own your data with Mastodon!