At a glance, one could be fooled into thinking the images were real.
The fabricated images — which had been viewed more than four million times — also depicted a scenario where the former US President broke free from cops and bolted, as a fake Melania Trump and Donald Trump Jr. shouted in protest.
Several of the disturbingly realistic images were shared widely on Twitter, with some falsely claiming they were legitimate.
The fakes came as a Manhattan grand jury weighed whether to indict Trump in connection to hush money paid to adult entertainer Stormy Daniels in 2016.
Eliot Higgins, founder of the investigative publication Bellingcat, tweeted out the deepfakes and said they were created with the text-to-image generator Midjourney.
The images were made using the prompt “Donald Trump falling over while getting arrested. Fibonacci Spiral. News footage,” he said.
“The Trump arrest image was really just casually showing both how good and bad Midjourney was at rendering real scenes, like the first image has Trump with three legs and a police belt,” Higgins said later, adding he didn’t think observers would take the shots seriously.
“I had assumed that people would realise Donald Trump has two legs, not three, but that appears not to have stopped some people passing them off as genuine, which highlights that lack of critical thinking skills in our educational system.”
As of Thursday, Trump has not been arrested.
Artificial intelligence saw another major leap this week upon the release of GPT-4, a more advanced version of OpenAI’s mind-boggling chatbot.
There are already fears for how education systems will be able to operate, with the chatbot now able to write exemplary essays on just about any topic in a matter of seconds.
As OpenAI co-founder Sam Altman explained when GPT-4 was announced on Wednesday, it is the “most capable and aligned model yet” with the ability to use images as well as text.
These are some of the jaw-dropping things people are already doing with the new tech.
Designer Pietro Schirano explained how he was able to recreate the game of Pong in under 60 seconds – on his first try.
“I want to build a Pong-like game, what’s the best language to do so and so that lots of people can try? Also I want to use Replit to built it,” Schirano wrote to Chat-GPT.
The AI recommended that he use the HTML5 Canvas API to create a game that runs directly in the browser.
“Here’s a basic example of a Pong game using JavaScript,” GPT-4 responded, amazingly delivering a string of code and a working version of the famous arcade game that was first developed in 1972.
AI engineer Sudharshan Chandra Babu took a photo of his fridge and asked GPT-4 to come up with ideas for recipes.
Babu first asked Visual-ChatGPT to describe in detail what was in the fridge based on the image.
He then commanded, “Use the items to come up with five food items that can be created with them. If prompted for more information, provide a quick recipe.”
GPT-4 responded, “Based on the provided ingredients, here are five food items that can be created”, and listed fruit salad, cheese omelette, ham and cheese sandwich, fruit smoothie, and cheese and fruit platter, along with instructions on how to create the meals.
Incredibly, one savvy user was able to “trick” the artificial intelligence into pretending it was a “human trapped inside a computer”, and successfully had it write lines of code to run on his machine, enabling it to “use it for its own purposes”.
“I think that we are facing a novel threat: AI taking control of people and their computers,” Michal Kosinski, a computational psychologist studying artificial intelligence.
“It’s smart, it codes, it has access to millions of potential collaborators and their machines. It can even leave notes for itself outside of its cage. How do we contain it?”
Spooky.
