Posts by jeremybmerrill@journa.host
(DIR) Post #AUjHyqxg3cEG1PDigS by jeremybmerrill@journa.host
2023-04-17T02:06:18Z
0 likes, 1 repeats
Dominion v. Fox News trial delayed to Tuesday, to allow conversations about the possibility of a settlementhttps://www.washingtonpost.com/media/2023/04/16/fox-news-dominion-trial/
(DIR) Post #AVupSSqc7QsvlIS7Lk by jeremybmerrill@journa.host
2023-05-22T13:50:38Z
0 likes, 1 repeats
Why are LLMs trained by having human beings rate proposed answers (RLHF), instead of just with regular supervised learning?@yoavgo@twitter.com's answers: it lets responses be more diversely-worded and lets the model learn how to say it doesn't know.https://gist.github.com/yoavg/6bff0fecd65950898eba1bb321cfbd81
(DIR) Post #AW1mTy4oYWgD9Oq7OK by jeremybmerrill@journa.host
2023-05-25T16:27:23Z
0 likes, 1 repeats
Facebook still appears powerless to stop bad guys from advertising a fake AI marketing assistant that's actually malware. The baddies have transitioned to impersonating Facebook parent company Meta... and Facebook still isn't acting.I first wrote about this problem in mid-April: https://www.washingtonpost.com/politics/2023/04/12/fake-chatgpt-preys-facebook-users/
(DIR) Post #AWUOPcxBuc1G3HQ8Z6 by jeremybmerrill@journa.host
2023-06-08T18:04:50Z
0 likes, 0 repeats
@ken hey who are you calling a nerd?!!
(DIR) Post #AWUd0D9mBcDY7Ie6am by jeremybmerrill@journa.host
2023-06-08T20:48:19Z
0 likes, 0 repeats
@ken phew I was worried you thought I was too much of an athlete to like the 990 data!!
(DIR) Post #AWtPs0yyI0jYSMLwzQ by jeremybmerrill@journa.host
2023-06-20T17:18:26Z
0 likes, 0 repeats
I got a Facebook ad for "real unbiased news written by AI""no reporters. no fake news."😂 where do I even start.
(DIR) Post #AXG6PytxTpoaF8ctua by jeremybmerrill@journa.host
2023-07-01T18:14:59Z
0 likes, 1 repeats
Is today the day to talk about the countermeasures that Facebook's anti-scraping team took to block the participants in the ProPublica Facebook Ad Collector project from sharing the ads they saw?All that effort, and I think we had a week or two of downtime over several years.
(DIR) Post #AXG6Q0q8H1wWFqqsE4 by jeremybmerrill@journa.host
2023-07-01T18:17:30Z
0 likes, 0 repeats
One time (on my birthday in 2018), they launched non-deterministic CSS for the "Sponsored" tag we used to preliminarily distinguish ads from non-ads. I distinctly remember my heart dropped when I discovered it, sitting on the patio at Golden Drops coffee shop on Clairemont.That was a mindfuck -- but ultimately not that hard to work around.
(DIR) Post #AXG6Q2XPxaA3WO6xKi by jeremybmerrill@journa.host
2023-07-01T18:23:08Z
0 likes, 0 repeats
Later, Facebook started to filter clicks on the menu containing the "Why Am I Seeing This" button that triggered fetching targeting data (e.g. ad limited to people 59+ who like Sean Hannity, Glenn Beck or Mark Levin) based on a JavaScript attribute in click events that distinguishes mouse-driven clicks from JavaScript-derived ones. That's a filter that wasn't implemented even on the Like button!! (Go have your bot make fake Likes on a minor celebrity's page all you want, I guess?)
(DIR) Post #AXG6Q4Brog6we82m1I by jeremybmerrill@journa.host
2023-07-01T18:25:41Z
0 likes, 0 repeats
Finally, they started adding extra characters, hidden with CSS,in the Sponsored label that distinguishes ads from regular posts. So, it looked like "Sponsored" when you saw it, but "SpSonSsoSredS" was the text content of the element. Trick was, screenreaders would read it as gibberish... 😬😬I forget how we worked around that one, but we did.
(DIR) Post #AXGAKMYH2HqNfk3mWO by jeremybmerrill@journa.host
2023-07-01T19:13:10Z
0 likes, 0 repeats
@ken @ndhapple it is indeed so sigh.
(DIR) Post #AXdHoXTKovYo8tFLLU by jeremybmerrill@journa.host
2023-07-12T21:40:22Z
0 likes, 1 repeats
Alert issued after rabid beaver terrorizes family on Lake Lanierhttps://www.ajc.com/news/crime/alert-issued-after-rabid-beaver-terrorizes-family-on-lake-lanier/AJKKGXYJ5RF4FJSRFJGUGMIIEM/
(DIR) Post #AYOiSykrmfWgsrNlOS by jeremybmerrill@journa.host
2023-08-04T17:27:32Z
0 likes, 0 repeats
Anyone know of a history of how French lost its status as lingua franca in favor of English? It seems like it would've been inconvenient to switch, especially for people in commerce, so there must've been a lot of inertia.#linguistics #etymology
(DIR) Post #AaLH4yDUugL3MbMhc0 by jeremybmerrill@journa.host
2023-10-01T18:41:18Z
0 likes, 0 repeats
I've seen a lot more people using Claude instead of ChatGPT lately. Curious why?
(DIR) Post #AaLIriOCeCYiQ5BCXQ by jeremybmerrill@journa.host
2023-10-02T00:18:38Z
0 likes, 0 repeats
@grammargirl got it, ty!
(DIR) Post #AaMwc1e3Thvapbcz7Q by jeremybmerrill@journa.host
2023-10-02T15:37:42Z
0 likes, 0 repeats
Which "champagne for my real friends, real pain for my sham friends"-style toast is better?1. ChatGPT's "Confetti for my loyal friends, bitter truth for my deceitful friends"Or2. My algorithm's: "focus for my real friends, real cuss for my faux friends"https://jeremybmerrill.com/documents/champagnerealpain.html
(DIR) Post #AcxRHoxb105LRmZzma by jeremybmerrill@journa.host
2023-12-19T03:56:37Z
0 likes, 0 repeats
@xek @simon Same error, actually. (Likewise llm 0.12, llm-llama-cpp 0.3) If it's helpful, `llm models` also doesn't show any gguf-related output.
(DIR) Post #AcyL8qQ1okyEnWAuP2 by jeremybmerrill@journa.host
2023-12-19T16:00:04Z
0 likes, 0 repeats
@simon @xek from `$ vllm plugins` I get `llm-llama-cpp` and `llm-gpt4all` (installed separately)xek's suggestion of `$ llm llama-cpp models` returns `{}` but fixes the problem and I can now run `llm -m gguf ...` (well, I run out of RAM, but close enough and I should've anticipated that)
(DIR) Post #AktayYSyGtyV44PtlA by jeremybmerrill@journa.host
2024-08-12T15:53:32Z
0 likes, 1 repeats
Almost 5 percent of Elon Musk's tweets consist of only the “tears of joy” or “rolling on the floor laughing” emoji.When reporting this story about Elon Musk's political turn to the right, I turned to one of my favorite natural language data exploration techniques: what are the most common strings in this dataset?So here you are: Elon Musk's most common tweets:😂🤣!!!YeahTrueWowExactlyGift Link: https://wapo.st/3WK2Ew8#datascience #nlp #datajournalism #data
(DIR) Post #AktayaD5muSgTP0FHs by jeremybmerrill@journa.host
2024-08-12T15:54:03Z
0 likes, 0 repeats
What's your first move when you open a new natural-language dataset? What fun or insightful stuff have you found?#datascience #nlp #naturallanguageprocessing #datajournalism #ddj #data #analytics