OpenAI Sora: Wake up Humans, Brace for Change
Less than a year ago, AI-generated video was considered a joke; now, look at the advancements. OpenAI’s latest model, Sora, transforms descriptive text into video, marking a significant leap forward that is nothing short of astonishing.
Sora generates “realistic and creative scenes” from textual prompts. Users can input detailed descriptions of scenes, and Sora will generate high-resolution video to correspond with them. This functionality aligns it with many other video generators developed over the past year or so.
Tracking the Evolution of AI from March 2023 to Mid-February 2024
To illustrate the advancements in this field, consider the developments from March to April 2023, as well as Google’s Lumiere system introduced last month. Now, observe OpenAI’s progress in mid-February 2024 with its latest Sora system, highlighting the remarkable pace of advancement. Below are several examples along with their corresponding prompts.
Accuracy and Detail in AI-Generated Scenes
Impressive. In general, the physics in these scenes are remarkably accurate. The details and movements are so realistic that one could easily mistake many for genuine footage if not scrutinizing for errors or realizing that it’s authentic-looking footage of something entirely fictional.
OpenAI also mentions that Sora can store characters, settings, and styles for future use in various scenes. This hints at its potential to eventually create complete narratives, television shows, or films.
Conversely, there is ample opportunity for enhancement, and like all AI systems geared towards creativity, the outcomes can be hilariously bizarre, particularly when requesting something exceptionally absurd.
And occasionally, it can produce an unpredictably artistic delight or two.
OpenAI’s Approach to Red Teaming with Sora
OpenAI is currently subjecting Sora to “red teaming,” where it’s exposed to inappropriate prompts to identify and prevent any potential misuse. While this process may impair its performance, it aims to enhance Sora’s compliance with copyright regulations and reduce the likelihood of generating objectionable content, making it safer and more suitable for family-oriented use.
But my goodness, in just a few months, we’ve progressed from this:
Evolving Toddlers or Accelerating Revolution?
These AI systems are akin to toddlers, exhibiting remarkable abilities yet still in the early stages of development. Their rapid progress surpasses that of any other technological field. It’s uncertain whether we’re approaching a plateau in advancement, followed by a gradual refinement process, or if we’re merely at the outset of an exponential growth curve in AI capabilities.
Some speculate that AI may reach its peak soon, leading to a decline in quality as it saturates more of the internet. Nonetheless, one thing is certain: humanity must embrace change, as the AI revolution presents unprecedented transformation. Unlike previous innovations such as the wheel or the internet, AI proliferation poses unique challenges to our societal structure and our position in the world. We are navigating uncharted territory, and the future remains uncertain.
Read the original article on: New Atlas
Read more: OpenAI´s Humanoids doing Great