3 ms·
Developer here. Like a lot of programmers, I struggle with distraction, and I get frustrated with time wasted on addictive social media. I also feel frustrated
by jedwhite 8y ago
Developer here. Like a lot of programmers, I struggle with distraction, and I get frustrated with time wasted on addictive social media. I also feel frustrated by the lack of control I have over the algorithms that control what I see on social media, and the constant tracking and privacy invasion.
So I've been working on Orac as a way to get interesting and useful content that matters from social media, without losing focus. It's currently an early preview version. The front end is built as a web app with React and GraphQL, and it uses deep learning to rank quality and predict attributes about stories it finds shared on social media (such as seriousness, objectivity, political stance etc). The back end is AWS Kinesis for stream ingestion, Lamda for running inference, and DynamoDB and ElasticSearch, along with AWS AppSync.
While it's experimental, the content predictions are already pretty interesting, and it has a bunch of pre-built filters as a way to play around with them - such as moods, personas and filter bubbles.
Would love to know what folks think and if this is something useful to people that's worth pursuing!
- fuddle 8y agoAwesome concept! An interesting use of AI. Do you think it would it be possible to recognize a news story as fake news? The dataskeptic.com podcast have been doing a series recently on fake news.
- jedwhite 8y agoThanks! If you watch the "Show Me Everything" raw feed for a while, it puts a "flag label" on different types of bad content - including clickbait, fake news, hate speech, extremism and state news and propaganda. A lot of those stories get stripped out when the filters are on. But the "crap detection" filter does flag those stories in the raw feed in real-time too. It's still learning but it's already decent at detecting this sort of content. I was thinking of adding a "fake news" view (or "Crap detector view") as it's interesting to see the techniques the people who propagate that stuff use!
- nikolay 8y agoIs this branding related to Blake's 7?
- Jaruzel 8y agoProbably not. The author seems american; The site has a very US news bias, and only has US politics filtering. Which is a shame. I hope the author plans to expand it to have a more global viewpoint. For those that don't know - 'Blake's 7' was a 1970s UK sci-fi TV show that had a portable 'super-computer' called Orac, whose personality was arrogant and aloof. The show was set in a dystopian future where a despised militarised government rules the galaxy. Still worth watching even now to be honest.
- nikolay 8y agoThanks for the summary. Blake's 7 was my favorite Sci-Fi show as a child.
- jedwhite 8y agoPart of Orac's mission is to be very judgmental about content when ranking its quality and importance, so the allusion to being aloof and arrogant is little bit of a joke, but purposeful :)
- jedwhite 8y agoI'd commented above that the current feed is intended for testing, but that the plan is for you to add your own social feeds and news sources, and that it will then skew similar in finding additional sources. The "filter bubbles" are also intended as a couple of examples, so initially I've used US ones. I'm Australian and while I'm based in the US, one of the problems Orac is trying to solve is localisation. Orac is running geo and locale predictions on content, but they aren't very good so far, so I haven't included them in the UI or filtering yet. But a few folks who I've shown have said that it would be useful to filter content that's relevant by location, so that is definitely in the roadmap!
- jedwhite 8y agoJust to clarify: that is a "yes this is intended for a global audience" :) The preview version is fairly US-centric (language used, initial sources followed, example filter bubbles etc). But that is very much a starting point to show how the concept works in action. So that will change as it develops. The underlying algorithms are built with both global use and local filtering in mind. So stay tuned on that front!