0 Followers
8 Following
7 Posts

@matthewcroughan @SecurityWriter Thank you for posting this. I was considering using OpenCode soon, and now I will have to take a close look first. "Home buyers" in the privacy policy - LOL.

Also the maintainer asking the evil question. While I get the "open source is not about you" point, that was quite a "finger on the button ready to silence you" moment of power

@TeensyTinyKobold @eff Yeah, I did jump on this one quote perhaps a bit too fast. My anger mostly comes from the negative parallelism, which is so overused by LLMs, being the *one thing* EFF picked out from the article. The mix of quote styles, shows something was copied, AI gen or not. When coding I've noticed AI overuses dashes in nouns that don't need them, and this article also has that ("start-up", "doubling-down", "Japanese-Americans"). Only the author knows the truth!
@mfru @eff I don't hate AI itself (I use it for coding help) but feel disrespected when I'm handed AI-generated text to read. I wouldn't have judged this line if it were not for the negative parallelism. These other typographical signs which weren't consistent throughout the whole article confirmed my suspicion.

@eff Please EFF, no more AI-generated content cowriting like this exact line you are quoting from the article 😢

Just this line has: 1) smart quotes, where standard quotes are used elsewhere in the article, 2) em dash, 3) "Not this, but that" negative parallelism (see linked Wikipedia article).

Please tell your writers we DON'T want to read AI-generated stuff.

https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing#Negative_parallelisms

Wikipedia:Signs of AI writing - Wikipedia

@stux I disagree. WinRAR is proprietary software that is great at making proprietary-format archives that are difficult to decompress using FOSS (unrar-free and 7-zip still stumble on a good portion of the format due to licensing restrictions). OpenAI has made a bigger impact on open research: fast, accurate transcriptions (whisper), RL (gym), and of course LLMs (starting with GPT-2, and of course ChatGPT and the "gpt-oss" models).