41–50 of 111 posts
Earlier quoted context omitted.
It works untill someone takes a screenshoot of the image
The watermark would still be there. You can take a photograph of the image with an old Polaroid and, if the resolution is high enough, the watermark would still be there.
> The concern is valid. But there’s a deeper worry, one that involves the enlargement not of our gullibility but of our cynicism. OpenAI CEO Sam Altman has voiced worries about the use of AI to influence elections, but he says the threat will go away once “everyone gets used to it.” > Some experts believe the opposite is true: The risks will grow as we acclimate ourselves to the presence of deepfakes. Once we take th…
It's the same idea as "lies spread faster than truth". Lies are often crafted to be especially juicy and salacious. Gossip has always been a problem; GenAI just extends this problem to other media.
Ars answered this much much better:
https://arstechnica.com/ai/2025/05/ai-video-just-took-a-star...
> As these tools become more powerful and affordable, skepticism in media will grow. But the question isn't whether we can trust what we see and hear. It's whether we can trust who's showing it to us. In an era where anyone can generate a realistic video of anything for $1.50, the credibility of the source becomes our primary anchor to truth. The medium was never the message—the messenger always was.
When something new is happening (or new information comes to light), and that thing has the potential to do harm, people come out of the woodwork to make doomsday predictions. Usually the doomsday predictions are wrong. A lot of these predictions involve technologies we all take for granted today. Like the telephone. People were terrified when they first heard about it. How will I know who's really on the other end?…
Even your example contains an unsolved, and serious problem. We still don’t know who is on the other end of the phone.
1) if a piece of content is a fact or not.
2) if the person you are acting with is a human or a bot.
I think its easier if you take the most nihilistic view possible, as opposed to the optimistic or average case:
1) Everything is content. Information/Facts are simply a privileged version of content.
2) Assume all participants are bots.
The benefit is that we reduce the total amount of issues we are dealing with. We don’t focus on the variants of content being shared, or conversation partners, but on the invariant processes, rules and norms we agree upon.
So We can’t agree on may be facts - but what we can agree on is that the norms or process was followed.
The alternative, to hold on to some semblance or desire to assume people are people, and the inputs are factual, was possible to an extent in an earlier era. However the issue is that at this juncture, our easy BS filters are insufficient, and verification is increasingly computationally, economically, and energetically taxing.
I’m sure others have had better ideas, but this is the distance I have been able to travel and the journey I can articulate.
Side note
There’s a few Harvard professors who have written about misinformation, pointing out that total amount of misinfo consumed isn’t that high. Essentially : that demand for misinformation is limited. I find that this is true, but sheer quantity isnt the problem with misinfo, its amplification by trusted sources.
What GenAI does is different, it does make it easier to make more content, but it also makes it easier to make better quality content.
Today it’s not an issue of the quantity of misinformation going up, it’s an issue of our processes to figure out BS getting fooled.
This is all putting pressure on fact finding processes, and largely making facts expensive information products - compared to “content” that looks good enough.
Earlier quoted context omitted.
Videos and photos have been faked for a long time. Nothing has changed in that regard but decreasing the effort required somewhat.
Your comment follows two persistent HN tropes: (1) ignoring the article, which deals precisely with why the cost of production matters, and (2) steadfastly refusing to recognise that quantity has a quality of its own - in this case a monumental reduction in production cost clearly leads to a tectonic reshaping of the information landscape.
Earlier quoted context omitted.
Logging them would be rather cost prohibitive, but images can be hashed + (invisibly) watermarked and video can be hashed frame by frame, in such a way that each frame authenticates the one before it. Surely there's a way to durably mark generated content.
I think such watermarks are one of DeepMind's goals: https://deepmind.google/science/synthid/
Earlier quoted context omitted.
Logging them would be rather cost prohibitive, but images can be hashed + (invisibly) watermarked and video can be hashed frame by frame, in such a way that each frame authenticates the one before it. Surely there's a way to durably mark generated content.
It works untill someone takes a screenshoot of the image
The problem then is repeated compressing causing that to occur inadvertently
When something new is happening (or new information comes to light), and that thing has the potential to do harm, people come out of the woodwork to make doomsday predictions. Usually the doomsday predictions are wrong. A lot of these predictions involve technologies we all take for granted today. Like the telephone. People were terrified when they first heard about it. How will I know who's really on the other end?…