The smarter AIs become, the scarier they can be.
When the world was caught by surprise, when the technology behind deepfake was released, people were awed. With AI, it's so easy to generate fake photos and videos.
Since the public began to realize about the technology's existence, the rest is history, but similar technologies that supplement "deepfake" were quickly developed.
And when OpenAI released its GPT-2, and later, the GPT-3, the technologies kept on branching.
Ultimately, the technologies made way for AIs like DALL·E and DALL·E 2, which are AIs that can produce original images from only text prompts.
And even scarier, was the more recent Stable Diffusion and Imagen AI, which takes things up a notch.
This time, it's Meta's turn, and it's nightmarish.
This is because Meta's AI can create videos using text prompts only.
Calling it the 'Make-A-Video' AI, the project was announced by Meta's founder and CEO Mark Zuckerberg himself.
"Make-A-Video solves this by adding a layer of unsupervised learning that enables the system to understand motion in the physical world and apply it to traditional text-to-image generation."
And according to Make-A-Video's website:
Initially, Make-A-Video offers three different styles of videos: surreal, realistic, and stylized.
And this in turn allows users to turn textual prompts into short videos, easily and quickly.
A small domesticated carnivorous mammal with soft fur, a short snout, and retractable claws
A golden retriever eating ice cream on a beautiful tropical beach at sunset, high resolution
A musk ox grazing on beautiful wildflowers
A blue unicorn flying over a mystical land
Humans building a highway on mars, highly detailed
With the developments of AI-powered text-to-image generators have been in full throttle. text-to-video has also become a thing.
And as for Meta, the company that is considered among the largest tech companies the world has ever seen, is revealing how fast the technologies have developed.
Since AI is a sensitive subject to some, and Meta is not known for being a privacy advocate, Meta said that is "committed to developing responsible AI and ensuring the safe use of this state-of-the-art video technology" to reduce the creation of harmful, biased, or misleading content.
The company said that it sources its data from millions of pieces of data it used so the AI can "learn about the world."
But before feeding the AI with the data, Meta said that it filtered the data "to reduce the potential for harmful content to surface in videos."
Meta also said that all videos made by Make-A-Video have watermarks to "help ensure viewers know the video was generated with AI and is not a captured video."
A small domesticated carnivorous mammal with soft fur, a short snout, and retractable claws
A golden retriever eating ice cream on a beautiful tropical beach at sunset, high resolution
A musk ox grazing on beautiful wildflowers
A blue unicorn flying over a mystical land
Humans building a highway on mars, highly detailed
And lastly, Make-A-Video is a work in progress.
Meta plans to make this technology available to the public. But before that happens, Meta wishes to "continue to analyze, test, and trial Make-A-Video to ensure that each step of release is safe and intentional."
Not long before this, the social media app TikTok introduced its own text-to-image generator.
While it's different to Meta's text-to-video, it looks like Meta is making something new, rather than blatantly copying others' features it's already known for.
