It now seems very plausible to me that at least one actor is willing and able to create and spread deceptively real deepfakes (e.g. Russia sabotaging the West). Why doesn't this happen all the time? Sure, there are a few here and there, but given that someone should be able and willing to do it much more often, where are the deepfakes? Maybe the problem is that a random video somewhere on the internet doesn't reliably go viral and that videos of politicians are effectively only seen and disseminated when they come from large media platforms? But the effect would have to be really strong to explain why so few deepfakes go viral.

This seems relatively important to me because it's a case of ‘as soon as enough actors have access to potentially dangerous technology X, some bad/careless actor will use it with bad consequences’. This argument is central to many x-risks. With deepfakes, I would actually expect the same, but it doesn't happen as much as I would expect. So there's some flaw in my model of how such apocalyptic residuals work, and I may be thinking about x-risks all wrong.

48 Reactions

New Answer

New Comment

4 Answers sorted by
Top

Charlie_Guthmann

Mar 03, 2025

Are you sure the technology is there yet? I don't know if I have actually seen an extremely convincing deepfake.

Karthik Tadepalli12d2

I consider myself good at sniffing out edited images but I can't spot any signs in Balenciaga Pope. Besides, for a deepfake to be useful, it only has to be convincing to a large minority of people, including very technologically unsophisticated people.

Charlie_Guthmann

12d

ok if we are talking about that level of deepfake I feel like that does already existed and has even existed before AI tools. "Fake News" is not an AI specific phenomena, though it makes it easier. Photo shop and captioning photos with lies do exist and go viral all the time. Boomers on facebook are fed a diet of bullshit. It's possible op doesn't use FB type sites and/or the sites have identified they don't like bs so that isn't what the algo feeds them.

Jonas Hallgren

Mar 04, 2025

I guess a random thought I have here is that you would probably want video and you would probably want it to be pretty spammable so you have many shots at it. Looking at twitter we already see like a large amounts of bots around commenting on things which is like a text deepfake.

Like I can see in a year or so when SORA is good enough that creating a short form stabel video is easy we will see a lot more manipulation of voters through various social media through deepfakes.

(I don't think the tech is easy enough to use yet for it to be painless to do it even though it is possible. I spent a couple of hours trying to set this up for a showcase once and you had to do some fine-tuning and training stuff, there was no plug and play which is probably a bottleneck for now.)

calebp

Mar 04, 2025*

Fwiw I think the "deepfakes will be a huge deal" stuff has been pretty overhyped and the main reason we haven't seen huge negative impacts is that society already has reasonable defences against fake images that prevent many people from getting misled by them.

I don't think this applies to many other mouse style risks that the AI X-risk community cares about.

For example the main differences in my view between AI-enabled deepfakes and AI-enabled biorisks are:
* marginal people getting access to bioweapons is just a much bigger deal than marginal people being able to make deepfakes
* there is much less room for the price of deepfakes to decrease than the cost of developing a bioweapon (photoshop has existed for a long time and expertise is relatively cheap).

J.K. Lund