this post was submitted on 09 Jul 2023
477 points (97.6% liked)

Memes

51640 readers
1574 users here now

Rules:

  1. Be civil and nice.
  2. Try not to excessively repost, as a rule of thumb, wait at least 2 months to do it if you have to.

founded 6 years ago
MODERATORS
 
you are viewing a single comment's thread
view the rest of the comments
[–] mob@lemmy.world 6 points 2 years ago (11 children)

Realistically, what are the dangers of AI?

I get that it might change employment around like I assume calculators and computers did, but what are the other reasonable things to worry about?

[–] Catsrules@lemmy.ml 10 points 2 years ago (9 children)

The biggest issue I see is the ability to create realistic fake content easily.

[–] CyanFen@lemmy.one 1 points 2 years ago (2 children)

And what is the danger in this? At this point everyone knows AI can make realistic fake content. It's unlikely that someone, say, in a position of power, would do anything rash after seeing an AI video considering the technology exists. No wars were started over photoshopped images.

[–] qisope@lemmy.world 4 points 2 years ago (1 children)

the target isn't people in power, the target of these tools is the general population. disinformation combined lack of critical thinking is already bad enough just with the posting of carefully cropped, edited, or out of context media, when the new tools can create realistic video with voice matched audio, more people will be fooled - plenty of them will happily believe whatever reinforces their existing position.

[–] Catsrules@lemmy.ml 1 points 2 years ago

Also if someone in power gets caught doing something bad they could muddy the waters saying it was AI and fake content.

[–] rockSlayer@lemmy.world 1 points 2 years ago (1 children)

Imagine the president giving a very important message to the people. Using content generation, a bad actor could insert minor alterations that change the meaning of an important sentence and then spread it naturally on social media. That could have dramatic implications for disinformation campaigns.

[–] mob@lemmy.world 1 points 2 years ago (1 children)

Could they not have done that 5-10 years ago?

[–] rockSlayer@lemmy.world 1 points 2 years ago (1 children)

Not nearly as well as it can now. Frame generation and speech imitators get better every day. Our AI is far better than it was 5 years ago, and 10 years ago algorithms like chatGPT and stable diffusion were things of science fiction.

[–] mob@lemmy.world 1 points 2 years ago

I mean it makes it more accessible to the general public, but for anyone was attempting to seriously deepfake a presidential announcement.. the resources have been out there.

Like 5 years ago, Jordan Peele did an Obama deepfake

It's not perfect, but it was pretty close. Someone with real motivation probably could have made it 100% believable

load more comments (6 replies)
load more comments (7 replies)