#disinfo
2024-08-05
chihao
13:21:08
AI-generated content farm (allegedly)
Ronnywang
昨天在聊奧運時,因為好奇保齡球有沒有成為奧運項目過,結果一 google 第一筆就寫「保齡球自1988年漢城奧運會時被列為表演項目,直到2024年巴黎奧運會才正式成為正式比賽項目。」,看了內容才發現那是個 AI 生成的內容的假訊息網站,專門透過 AI 生成內容來賺廣告流量費的
- Forwarded from #rand0m
- 2024-08-05 08:39:58
chihao
13:21:08
AI-generated content farm (allegedly)
Ronnywang
昨天在聊奧運時,因為好奇保齡球有沒有成為奧運項目過,結果一 google 第一筆就寫「保齡球自1988年漢城奧運會時被列為表演項目,直到2024年巴黎奧運會才正式成為正式比賽項目。」,看了內容才發現那是個 AI 生成的內容的假訊息網站,專門透過 AI 生成內容來賺廣告流量費的
- Forwarded from #rand0m
- 2024-08-05 08:39:58
2024-08-17
@null
11:28:13
Hi!
@null
11:31:05
I just tried to log in to Notion but prevented by bad internet connection (sad)
@null
11:32:38
Hi Johnson (MrOrz) from Cofacts here!
@null
11:32:45
Welcome!!
@null
11:34:43
I just want to share the blog entry (written in Japanese) which anaryzed CNs on media account post https://note.com/shioshio38/n/nf99e946ceed9
note(ノート)
後半一部有料です。論旨は無料で読めます。定期購読がお得です。 時間がない人向けまとめ マスコミのツイートに付いたコミュニティノートをまとめた。そのコミュニティノートをもとにマスコミ報道にどのようなミスリーディングな記述が多いかを確認した。 シンプルに誤報と言えるものから、重大なミスと軽微なミスを混ぜて件数を報道する、別の重い罪状を書かずに見出しを書くなど多様なミスリーディングを誘う記述があることが分かった。 マスコミ報道はSNS上の書き込みと違い政治家や企業幹部など国民に大きな影響を与える意思決定者の決定の素材になる。そのマスコミ報道がミスリーディングを誘うものであれば国益を損なう![]()
- 👍1
@null
12:19:05
<system> file fto_birdxplorer_data.zip too big to download (6135695 > allowed size: 1000000)
@null
12:20:49
You can run API server by yourselves 1. Clone https://github.com/codeforjapan/BirdXplorer 2. Go to /api directory 3. Run `docker-compose up -d app db` 4. Import data from csv
GitHub
Contribute to codeforjapan/BirdXplorer development by creating an account on GitHub.![]()
@null
14:13:19
This is embed api ↑
@null
14:20:02
Will use svelte kit, we will use [Svelte Kit API routes](https://kit.svelte.dev/docs/routing#server) for CORS proxy, so no need for CORS headers anymore
@null
14:47:53
@Jasmine This is environment vars info https://fastapi.tiangolo.com/advanced/settings/#settings-and-environment-variables
fastapi.tiangolo.com
FastAPI framework, high performance, easy to learn, fast to code, ready for production![]()
@null
14:59:06
So I’m waiting the search API comes!
@null
15:02:54
@null
15:35:47
We can extract URLs given a URL to twee. However, it’s extacting multiple URLs at once, including: 1. hashtags (with host being `[twitter.com](http://twitter.com)` ) 2. external URLs (host being `[t.co](http://t.co)`) 3. links to the media (images, videos, etc, also starting with `[t.co](http://t.co)`) https://github.com/codeforjapan/disinfo-fto/pull/4 I guess only the 2 (external URLs) are useful, but we can’t differentiate 2 from 3.
X (formerly Twitter)
From breaking news and entertainment to sports and politics, get the full story with all the live commentary.
GitHub
Add a scaffold for getDataFromTweetUrl, with a working mechanism that turns twitter URL to the URLs inside. TODO: invoke BirdXplorer API to get posts with these URLs Example input: <https://twitter>....![]()
@null
16:36:25
@osoken
@null
16:39:49
```curl -X 'GET' \ 'https://birdxplorer.onrender.com/api/v1/data/posts?offset=0&limit=10&search_text=https%3A%2F%2Ft.co%2FNkj305cUOb' \ -H 'accept: application/json'``` this request will return a single post
@null
16:58:33
GitHub
Contribute to codeforjapan/disinfo-fto development by creating an account on GitHub.![]()
@null
17:04:55
@null
17:06:33
This is `[notion.site](http://notion.site)` and maybe I don’t have write permission
@null
17:09:31
@null
17:09:39
Achivement Made Chrome Extension to warn CN existence on third-party news site https://github.com/codeforjapan/disinfo-fto/tree/main/chrome-extention (only frontend)
GitHub
Contribute to codeforjapan/disinfo-fto development by creating an account on GitHub.![]()
@null
17:26:26
Frontend App is now deployed https://disinfo-fto.vercel.app/
2024-08-18
@null
09:16:00
Post in BirdXplorer API that has no Note attached: ``` { "postId": "1798677544457380190", "xUserId": "20674741", "xUser": { "userId": "20674741", "name": "savechildrenuk", "profileImage": "https://pbs.twimg.com/profile_images/1572224248088780800/23kY9CR1_normal.jpg", "followersCount": 195130, "followingCount": 1931 }, "text": "We're horrified at yet another attack on a UN-run school in Gaza that has killed dozens, including children.\n\nWe're in a never-ending loop of killing, condemnations, and inaction.\n\nCivilians in Gaza are paying the heaviest price for world leader's inactions. \n\n#CeasefireNOW https://t.co/je6JdwTn7J", "mediaDetails": null, "createdAt": 1717673119000, "likeCount": 2145, "repostCount": 1348, "impressionCount": 682535 }```
X (formerly Twitter)
We're horrified at yet another attack on a UN-run school in Gaza that has killed dozens, including children. We're in a never-ending loop of killing, condemnations, and inaction. Civilians in Gaza are paying the heaviest price for world leader's inactions. #CeasefireNOW![]()
@null
09:29:20
I was trying to search using URLs from news sites through a Chrome extension, but it didn’t work well. I just figured out the cause of the issue. When I tried using a URL like `https://birdxplorer.onrender.com/api/v1/data/posts?offset=0&limit=10&search_text=asahi.com`, nothing came up. This made me realize that URLs like this are stored after being converted to `[t.co](http://t.co)`. The topic of URL canonicalization did come up briefly in a brainstorming discussion, but we decided to postpone it. To determine if a CN exists for URLs obtained from a Chrome extension linked to news sites, we need to address this issue. I’ll try the conversion of URL today.
@null
09:38:53
Convertion from `[t.co](http://t.co)` to original URL is easy ```>>> import requests >>> >>> def get_redirect_url(short_url): ... response = requests.head(short_url, allow_redirects=True) ... return response.url ... >>> print(get_redirect_url("https://t.co/Nkj305cUOb")) https://twitter.com/YE9DL/status/1798672472696635856/photo/1```
X (formerly Twitter)
20時30分から配ります 数時間限定ですが全員に配布します 但しまだもらえてない人限定です 数時間限定で反応してくれた人 に全員配布します pay pay 100000円 いいね、フォロー、リツイート それでは開始。 #PayPay配布 #PayPay希望![]()
X (formerly Twitter)
20時30分から配ります 数時間限定ですが全員に配布します 但しまだもらえてない人限定です 数時間限定で反応してくれた人 に全員配布します pay pay 100000円 いいね、フォロー、リツイート それでは開始。 #PayPay配布 #PayPay希望![]()
@null
09:41:57
After small survey. It looks no easy way to make given URL to `[t.co](http://t.co)` because it is X’s internal service and no API exposed. So we can not `convert given URL to [t.co](http://t.co) and search` strategy.
@null
10:11:15
X (formerly Twitter)
Another of Joe Biden’s culture enriching border hoppers has committed murder.
@null
10:14:39
I made tco-expantion script but because it is single thread, estimated 5 hours to process 😂 I’ll make parallel them https://github.com/codeforjapan/disinfo-fto/blob/main/script/expand_tco.py
GitHub
Contribute to codeforjapan/disinfo-fto development by creating an account on GitHub.![]()
@null
10:25:07
@りょーま(Yuki Kawabe) https://t.co/kMAG1sPvjv in `search_text` triggers internal server error
X (formerly Twitter)
NEW: ICE confirms that a Mexican man charged with the killing of two children in Wisconsin was in the country illegally. He entered legally in 2016 and stayed past his date of departure. <https://t.co/zkVdOXFFxi>
@null
10:31:51
@null
10:32:13
If it is acceptable to use a document in the API to just demonstrate [the UI](https://disinfo-fto.vercel.app/) is working, this URL would work: https://x.com/RepMikeCollins/status/1812203309568258254
@null
10:37:04
parallel conversion also takes time… `1751 / 11471 (305 sec)`
@null
11:00:27
Minor question: do we need a `LICENSE` file in https://github.com/codeforjapan/disinfo-fto?
GitHub
Contribute to codeforjapan/disinfo-fto development by creating an account on GitHub.![]()
@null
11:12:54
I noticed some url takes long time (e.g. 30 sec) because of server error
@null
11:14:27
Oops I killed the pocket-wifi :lol: ```% git pull --rebase fatal: unable to access 'https://github.com/codeforjapan/disinfo-fto.git/': Could not resolve host: [github.com](http://github.com)```
GitHub
Contribute to codeforjapan/disinfo-fto development by creating an account on GitHub.![]()
GitHub
GitHub is where over 100 million developers shape the future of software, together. Contribute to the open source community, manage your Git repositories, review code like a pro, track bugs and fea...![]()
@null
11:35:57
Samples of URLs need canonicalization https://github.com/codeforjapan/disinfo-fto/blob/main/script/reverse_map.csv ```https://edition.cnn.com/politics/live-news/election-biden-trump-07-13-24/index.html?Date=20240713&Profile=cnnbrk&utm_content=1720914448&utm_medium=social&utm_source=twitter https://t.co/tl4nWE2ujD https://edition.cnn.com/politics/live-news/election-biden-trump-07-13-24?Date=20240713&Profile=CNN&utm_content=1720909935&utm_medium=social&utm_source=twitter#h_872751b6f0b528130bbad5b797e424b1 https://t.co/kohybm4y5M https://edition.cnn.com/politics/live-news/election-biden-trump-07-13-24?Date=20240713&Profile=CNNPolitics&utm_content=1720909962&utm_medium=social&utm_source=twitter#h_872751b6f0b528130bbad5b797e424b1 https://t.co/jcyQryT1S8 https://edition.cnn.com/politics/live-news/election-biden-trump-07-13-24?Date=20240713&Profile=cnnbrk&utm_content=1720909856&utm_medium=social&utm_source=twitter#h_872751b6f0b528130bbad5b797e424b1``` (updated because the file was updated too)
GitHub
Contribute to codeforjapan/disinfo-fto development by creating an account on GitHub.![]()
@null
11:36:54
it is sorted list of all expanded URLs
chihao
11:46:28
\ disinfo hackathon in November /
yuhsuan660
12:11:23
@yuhsuan660 has left the channel
2024-08-19
@null
02:02:21
I processed all data in post.csv and pushed to the repos
2024-08-27
@null
11:03:54
The event site for the November hackathon is now live! The event will be conducted entirely in Japanese.🙇 https://www.disinfo2024.code4japan.org/
Code for Japan on Notion
Code for Japanは、日本初の偽情報をテーマにしたハッカソン「Hack the Disinfo 2024」を11月3日〜4日に開催します。![]()
- 🙌3