The Deep Fake Decade: when world leaders have no idea what is real or not real

In case you missed it, Vladamir Putin did a media conference the other day and it did not include just the press, but the general public too. Questions like “when will the war end”, “why does my mobile phone say that I’m no longer in the Russian network when travelling to Crimea” and “tell us, when will our lives get better?” are examples of what was asked.

The reason I’m blogging about this is that one “resident of St Petersburg”, a student, asked the Russian leader if it is true he has lots of body doubles.

The thing is that the questioner was Vladamir Putin in his own voice and with his own face. A deep-fake generated Putin. What I loved about this is two-fold: Putin’s face when he looks up and sees who is the questioner; and then the fact the questioner asks about how many body doubles does he have.

https://www.youtube.com/watch?v=q-IgEbKKG6Y

Putin responded and states: "I have decided the only one who looks like me and speaks with my voice should be me.”

This was like a photo bomb interjection in his conference, but it made me think that we can now see anyone, anywhere, anytime in any situation, created by a system that is out of control.

By way of example, there’s a very interesting investigation by Bloomberg into fake porn. A group of women who were at high school with someone they considered a friend discovered that he had taken their social media images, manipulated them, and then posted them on porn sites, showing their naked bodies.

There’s a “weird and creepy (web)site where someone is posting explicit photos of girls from school and writing about them being raped and murdered. ‘There’s pictures of you on it, and I wanted you to know that’ …

“The phone buzzed a link to the website: cumonprintedpics.com. Luque started scrolling. She saw a photo she recognized. It was her, at 18, standing in a dressing room. But the swimsuit she’d been wearing in the original photo, the one she’d uploaded to social media, was gone. Someone had digitally altered the picture so it looked like she was posing completely naked. She knew the breasts weren’t hers, but they looked real enough that other people might think they were. She was too stunned to speak.

“The next image made her gasp. It showed a printout of a photo taken when she was 5, with the chubby cheeks and ringlets she’d long since grown out of. An erect penis rested atop the photo, touching her face. The accompanying post encouraged men to ejaculate on it.”

This is why the European Union has led the world in drafting AI regulations but, as The Economist highlights, it’s a bit of a mixed bag. Over a 40 hour long meeting earlier this month, the Eurocrats managed to agree new laws on the use of AI in a process that started five years ago. The thing is that, having started five years ago, it was well before ChatGPT and other Large Language Models (LLMs) and now Large Multimodal Models (LMMs)*.

In other words, the five year cycle to generate regulations started with a target towards how AI was being used in chatbots and things, and now has to deal with a whole new generation of AI. That’s why it’s a mixed bag.

Other countries – Brazil and Canada – are also drafting AI regulations, but the real question is how such regulations would deal with the situation highlighted above by Bloomberg?

During their reporting, it turned out that there were no regulations related to deepfake porn. They did finally manage to bring the perpetrator to court however. This is because he had used photographs of children as porn, and that breaches the laws related to underage sex.

Either way, we live in a time of massive change where even world leaders have no idea what is real or not real.

Any relevance to banking and finance? Well, obvs, we can defraud everyone, everywhere, every time, all of the time. More on that later ...

 

* LLM relates to text; LMM relates to multimedia