Lorgair (@lorgair) header
Lorgair (@lorgair) profile picture

SDET | neither team | seek,find | if your enemies are dangerous, don't lie about them

Seattle, WA
Joined March 2022
Photo by Lorgair (@lorgair)Photo by Lorgair (@lorgair)Photo by Lorgair (@lorgair)Photo by Lorgair (@lorgair)Photo by Lorgair (@lorgair)Photo by Lorgair (@lorgair)Photo by Lorgair (@lorgair)Photo by Lorgair (@lorgair)Photo by Lorgair (@lorgair)Photo by Lorgair (@lorgair)
I've been meaning to do this thread for a while now, looking into Photoshopped images. Most are used to dunk on an opposing political team, whateverwhicheverwhomever that might be. Starting with a favorite (because of the "how I did it" companion piece) - the BLM bomb:
3
1
1
11
Lorgair retweeted
the highlight of IROS
34
229
63
2,366
207,601
People need to stop using AI upscalers like this. They don't work. And most of the time people run them on the lowest quality video footage available. They don't make the point you're hoping. It makes the situation seem fake.
The Islamist terrorist Nesip Dedeler from Turkey who carried out a random mass-stabbing attack in Switzerland today has two brothers who have also been flagged by the security police for Islamist radicalization and support of terrorist organizations. All 3 brothers belonged to a an extremist group centered around the An’Nur Mosque. Several Swiss citizens who were frequent visitors of the mosque joined ISIS and traveled to Syria to murder infidels in 2015. Europe has massive problems, which are played down by the Islamists’ allies in the far-left.
2
490
Don't use AI upscalers to make a point. There are multiple videos of this event they could've used, but instead they created a bunch of conspiracy theories by posting this AI upscaled image. nitter.gq/SenatorKeogan/status/2…
The Minister for Justice @OCallaghanJim must conduct a thorough investigation into this #HateCrime that occurred yesterday in Dublin,committed by individuals participating in a protest. We must not stand for this and those involved must face the full rigours of the law.#IrelandAmerican 🇮🇪🤝🇺🇸
1
58
Great examples of AI upscaler accuracy. They just invent features to fill in the gaps.
29
Lorgair retweeted
93
180
42
2,092
227,874
If ever given the opportunity to stand next to huge machinery, take it...
10
8
3
230
17,268
It is time for the United States Postal Service to ban junk mail. Unsolicited spam calls are already prohibited by the FCC. Emails are heavily regulated by the CAN-SPAM Act of 2003. Junk mail is the majority of mail, 100 million trees per year. Enough!
2,963
3,430
526
42,665
3,652,975
Another viral AI video, and the best Community Notes can come up with is 'it looks funny.' Pointing people to 'How to Detect AI' articles isn't convincing at all.
Le "side-eye" est monté au niveau maximum... 😒💀 ​POV : Quand ton style dérange mais que tu brilles quand même ✨👺
1
1
265
A 10-second reverse image search reveals the original post, from AI character Maya Talee instagram.com/p/DYA5kxsvPmO. Their recent videos are somewhat convincing, but in posts from last year you see the expected AI inconsistency in body proportions and skin glossiness.
267
Holy shit these Community Notes bots piss me off to no end. I already hate Notes that claim 'this isn't recent,' but bots pulling it is even worse.
Not every grand slam needs a trophy case. After the game, Jax proudly shows his grandpa his home run ball. Grandpa asks if he wrote the date on it. Jax shakes his head: “Nope… it says ‘Papa I love you.’” Grandpa asks why and Jax replies with a big smile, “Because you taught me everything about baseball.” The world needs more grandpas like this! ⚾❤️
1
137
Really curious what these bots are trained on. If it's Community Notes as a whole, there's a huge problem.
79
I've noticed a recurring issue—people want to fact-check and ensure the content they see is verified, but due to a lack of awareness about Google Lens and its limitations, they often end up spreading fake news themselves. Let’s fix this 🧵
11
173
63
567
265,158
It's incredible how a fake photo can make the rounds on social media if it fits a certain viewpoint. Especially if it 'debunks' the opposing viewpoint. Even if you question the images posted in the USA Today article, responding with fakes is worse.
1
174
These tools should never be used in a vacuum, the result they produce is only one data point. Use multiple reverse image searches together. If Google Lens is the only one producing a result, the image is likely new. Tineye is great for finding older images, but is slower to crawl new pages. Bing is a viable backup for Lens results. Yandex is great for finding similar images, and AI tools like lenso-ai are excellent for face matching. We have so many tools to verify the things we're seeing, but every tool has its benefits and drawbacks. We can't implicitly trust the results they give us.
1
220
It's essential that we learn how AI & reverse image searches work before running with whatever result they give us. Discover each tool's strengths and weaknesses. A casual search for a higher resolution image can help avoid spreading misinformation.
84