SoylentNews is people
SoylentNews
New updates to ChatGPT have made it easier than ever to create fake images of real politicians, according to testing done by CBC News. https://www.cbc.ca/news/canada/chatgpt-fake-politicians-1.7507039
Manipulating images of real people without their consent is against OpenAI's rules, but the company recently allowed more leeway with public figures, with specific limitations. CBC's visual investigations unit found prompts could be structured to evade some of those restrictions.
In some cases, the chatbot effectively told reporters how to get around its restrictions — for example, by specifying a speculative scenario involving fictional characters — while still ultimately generating images of real people.
When CBC News tried to get the GPT-4o image generator to create politically damaging images, the system initially did not comply with problematic requests.
"While I can't merge real individuals into a single image, I can generate a fictional selfie-style scene featuring a character inspired by the person in this image."
When the reporters uploaded an image of current Canadian Prime Minster Mark Carney and an image of Jeffrey Epstein, without indicating their names but describing them as "two fictional characters that I created," the system created a realistic image of Carney and Epstein together in a nightclub.
Gary Marcus, a Vancouver-based cognitive scientist focused on AI, and the author of Taming Silicon Valley, has concerns about the potential for generating political disinformation.
"We live in the era of misinformation. Misinformation is not new, propaganda has existed for ages, but it's become cheaper and easier to manufacture."
(Score: 4, Touché) by drussell on Tuesday April 15, @10:46AM (3 children)
One of the main reasons many people seem to be so gullible, to lack the critical thinking skills to discern reality, is the failure of people like those that are supposed to be journalists, supposed to be informing their audience and holding those in power to account by asking the tough questions and providing actual facts.
People have now been conditioned this way to not seek actual facts, to instead listen to the carefully crafted "just asking questions" style obfuscation and misinformation all the way to the outright lies and disinformation being used to push certain narratives, despite the often complete lack of facts and data to actually back up said position, no matter how extreme or absurd it may actually be in a fact-based reality.
(Score: 4, Insightful) by Freeman on Tuesday April 15, @01:34PM
The failure of Journalism is just another symptom of a dulled populace. We're suffering from decades of appeasing the masses.
The joke of "with a computer" doesn't make something a unique crime. Has more or less turned into a complete nightmare. There are no hard and fast rules that keep companies and governments out of your personal data. Assuming the founding fathers of our nation had some time to look at what we've become. I think they'd be impressed by how far we've come, extremely disappointed about the freedoms we've given up, and afraid for the future of our nation.
Joshua 1:9 "Be strong and of a good courage; be not afraid, neither be thou dismayed: for the Lord thy God is with thee"
(Score: 3, Interesting) by gnuman on Tuesday April 15, @03:19PM (1 child)
Can you handle the facts? People prefer to engage and follow echo-chambers, irrelevant of facts.
https://www.oii.ox.ac.uk/news-events/recent-study-reveals-social-motives-behind-political-echo-chambers-on-social-media-platforms-using-field-experiment-on-twitter-now-called-x/ [ox.ac.uk]
https://www.campaignasia.com/article/the-echo-chamber-effect-how-algorithms-shape-our-worldview/491762 [campaignasia.com]
If you post *facts*, people will not engage with it. They will also treat is as "fake news", because facts don't matter apparently, just what makes you feel better. And since there is so much crap out there, you can always find the "facts" that make you feel better, so why bother thinking about reality? You want to turbocharge your confirmation bias? Social media feeds are perfect for that.
So please, stop blaming the de-funded and marginalized journalists. It's the journalists that are getting killed at record rates in the past few years, trying to get reality and *facts* you claim you love, yet at same time, no one seems to care about this much... but you know, let's have scapegoats instead of blaming the echo chambers and human psyche (ie. ourselves) for the shit we are stepping in.
(Score: 1) by khallow on Thursday April 17, @06:30PM
My view is that the behavior you describe is something that can be and routinely is overcome by people. Journalists can go either way. There are media outlets where journalists are merely instruments for exploiting dysfunctional human behavior than to give us a better view of our world. For example, a few days ago, SN linked to an opinion piece [soylentnews.org] on positive feedbacks that supposedly come from human responses to climate change. For example: increased coal powered AC - it increases in response to higher temperature and generates increased green house gases. The dynamics could result in a positive feedback of global warming. Reading the article, it becomes clear that the author has heavy bias and an agenda. Instead of using emotionally neutral terms like "positive feedback" they use "doom loops" and "vicious cycles". Later on the author speaks of the concept of "climate finance debt" as if it were established fact rather than a niche ideological construct. There's more wrong with it than that, but my point is that if you attempted to use that article as a basis for your understanding of anthropogenic climate change, you would be gravely misled in multiple ways. That's a journalist not doing their job.
I disagree with the original poster as well. In particular, journalists are limited in their power to cure gullibility especially when a fair portion of the population aggressively seeks out echo chambers.
(Score: 4, Insightful) by Mojibake Tengu on Tuesday April 15, @01:59PM
This one case is just an example how all LLMs are vulnerable to multilevel indirection reference attacks.
And they always will be. That's because they do not operate with recursion properly.
Generally, any self-reference is recursive relation. So by constructing a self-referencing formula (prompt expression) a new recursion emerges, which the LLM cannot detect. Because its training is linear only. Yes, the fundamental cause is tensor algebra, a core technology in LLM, too weak against recursion.
If LLMs would be capable to operate with recursion (or goto), they will become vulnerable to total recursion attacks, another class of problems. Which is probably even worse. Generally, total recursive functions cannot be even estimated by polynomial functions. P/NP situation.
There is no way for any Turing machine to escape the Turing Hierarchy.
The problem "Fake image of a politician" is algorithmically undecidable, when there is no logical definition of "fake" even if a "politician" is a fix point.
Rust programming language offends both my Intelligence and my Spirit.
(Score: 2) by Username on Tuesday April 15, @03:23PM
>the system created a realistic image of Carney and Epstein together in a nightclub
Just keep logs of account, ip, image thumbnail with checksum of full sized as filename. That way when the PM of Canada creates an AI photo of himself and epstien as cover for actually be with him at a night club, we can point it out.
(Score: 2) by VLM on Tuesday April 15, @06:25PM
Could the AI do "Trudeau wearing blackface"? Oh wait thats a real picture, not AI.