Towards a synthetic reality

The inexorable and relentless progress of generative AI.

Date

July 19, 2024

Category

Reading time

10 min

A sign that we are approaching a “simulation singularity” is the steady progress in generative AI for creating images from prompts. Year after year (sometimes even faster), diffusion models like Midjourney, Firefly, Stable Diffusion, and countless others, continue to deliver higher levels of real-world similitude, approaching a level of perfection previously unimaginable. What’s next? Video is undoubtedly the next frontier. What happens when video is generated in real-time? How about fusing these images with 3D gaming engines like Unreal? Imagine an AI world simulation in high-resolution, 360º experiences (i.e., AR or VR). The Matrix is not far from today.

Sinthetic Reality COVER

Roma Goat

Trained goats perform in towns by nomadic Roma. Using an accordion and other instruments, the animal handler directs the goat to perform feats of equilibrium on chairs or similar contraptions. The noise attracts neighbors who would throw some coins in appreciation.

December 2022

These images were created on December 12th, 2022, using DALL-E 2 (Which was released to the public on April 6, 2022). As the first generative image engine by OpenAI (the creators of ChatGPT), DALL-E sparked excitement with its potential to turn any word into an image. Note the lack of realism in the details, with an composition resembling a collage. Nevertheless, I remember being quite excited on its "potential", and thrilled how my written thoughts could conjure an image.

El marrano de San Antón

Every town has a free pig that roams the streets (“El marrano de San Antón”) and is collectively fed by all its neighbors. At the end of the year, it will be sacrificed and enjoyed during the festivities.

July 2023

These images were created on July 18th, 2022, using Midjourney 5.2. The leap of progress just six months from previous version was astonishing. Images are now useful in professional work (Generative AI is no longer a toy)... although perhaps a bit too polished. The overall feeling was borderline Dysney, too perfect... People are almost like CGI dolls. but the amount of detail extrapolated from a simple paragraph indicated that synthetic reality was on its way.

Watermelon Truck

Traveling salespeople offer watermelon slices to kids in each small town: “For 25 pesetas, you can eat, drink, and wash your face.”

July 2024

These images were created on July 18th, 2022, using Midjourney 6. Notice how much more detailed is the hair, the clothing texture, the real skin, vehicle surfaces, complex lighting, and with obvious improvements in overall layout, lighting, and framing (quite the professional "journalistic" aesthetic).

Town Crier

The official town crier delivers weekly news from city hall with a megaphone on different spots in town, waking neighbors early in the morning.

Honorable Pickpocketing

Pickpocket thieves had the honor code of returning the wallet (minus the cash) to any public mailbox with all the identification documents still within. The national mail service had a service to make sure it arrived back to the wallet’s owner.

Flies' Reflection

A street in a small town in Spain features nylon-hanging plastic water bottles on each door to “scare off” flies when they see their own reflection.

Charred Potatoes

Every October, people in the town of Carcelén (Spain) make big bonfires, roast potatoes until charred, and rub them on each other’s faces as a harvest ritual of bonding and celebration known as “El Somarro.”

70's YOLO

Families in the 1970s crammed up to six kids or more into tiny European cars. There were no seatbelts, child car seats, or airbags.

Summary of the article

As generated images from AI reach their peak realism, we wonder what is next. Video no doubt (OpenAi's Sora demo technology has already shown us the potential) Real-time AR and VR environments will come next... then a fully simulated reality? Behold, the birth of The Matrix.