Latest / Situational Awareness Tactics / AI AND THE DANGERS OF DEEP FAKES2
Transcript
- 0:01Welcome to Situational Awareness Tactics podcast. This podcast provides the crucial art of understanding
- 0:09current elements in an environment to increase your safety and survival. Here's your host,
- 0:15forensic psychologist and consultant doctor Carlos Google CEO Sound of the Alarm, about
- 0:25the dangers of using AI to create deep fake videos of public figures, a
- 0:30task it says, it will be easier and easier. If you look a
- 0:32lot of the advertisements on Instagram, you'll see a lot of apps that can
- 0:37do that, and it is really remarkable. Some of them are great for
- 0:42marketing and doing promotional videos, but man, it's crazy. And they have
- 0:46avatars now, and so they're not really imitating other people, but some of
- 0:50them are imitating you. In other words, you can put yourself and create
- 0:53an avatar of you saying things only a matter of time where they create lons
- 0:58of other individuals and what he says, the concern is that you'll be able
- 1:03to create a video pretty easily of an advat with an avatar AI wise,
- 1:10and you could be saying something, or somebody else could be saying something,
- 1:12but you never said it, but in the video it looks like you actually
- 1:15said it, and it could look accurate and on a societal scale, this
- 1:19can cause a lot of harm. So when they asked the Google CEO about
- 1:25barred and it's safety for society, he says, this the way we have
- 1:29launched it today as an experiment, in a very limited way. I think.
- 1:32So it wasn't very reassuring, But we all have to be responsible.
- 1:37You know, good luck, there's millions and there's all millions of people who
- 1:38are not responsible. Yes, there's millions that are, but the millions that
- 1:44are not can cause enough chaos with it. And we know that because we
- 1:48did the podcast yesterday about the AI. The scammers using AI to extort money
- 1:53out of individuals by faking voices of their victims by grabbing them from social media,
- 1:57so that we have to be responsible is already out the window. One
- 2:04part of it is also that they have to get the user feedback. They
- 2:07can develop more robust safety layers before they can build. They need to find
- 2:09out what's going on out there. They those need to monitor it much more
- 2:13closely. Current AI models are already capable of fabricating images of public figures that
- 2:16I mentioned that are nearly in the swing distinguishable from reality. Now the video
- 2:23and audio fabrications are less adept so far, but they're getting better fast.
- 2:29AI created audio clips are capable of imitating the intended voice, like I mentioned
- 2:31for the scammers, but current versions sound slightly robotic, not really, not
- 2:36a lot of them. AI video clips are even less polished. The CEO
- 2:40went not to say that Google does not always fully understand the answers it's barred
- 2:45provides, offering an example of the program appearing to teach itself Bengali despite not
- 2:50being trained for it. That's bizarre. You don't fully understand how it works,
- 2:57and yet you've turned the loose on society, asked the porter. Yeah,
- 3:00let me put it this way, it says the CEO. I don't
- 3:02think we fully understand how a human mind works either. That's brilliant. So
- 3:07obviously you go back to the Jurassic Parks line. We're so busy wondering if
- 3:14we couldn't We didn't ever ask the question if we should, And now they're
- 3:17unleashing this thing supposedly an unlimited way. But what an answer really the carelessness
- 3:23of these individuals. So we'll see AI and how it does look. I'm
- 3:28an AI fan I think there's a lot of good things to it, but
- 3:30at the same time, it's a lot of dangerous to it. We saw
- 3:32that with social media. We'll keep an eye on this. That's it for now.