This week, both Apple and OpenAI debuted their public version of their new expansion of their AI multiverse. Let’s take a look at Sora and what it plans to bring to the AI video landscape.
Arguably, the internet’s most respected tech reviewer, Marques Brownlee, from MKBHD, has the best and most succinct breakdown of what Sora has to offer right now.
MKBHD Review
Cool! I’m a student/faculty and wanna try it out!
Before you get your hopes up, there is a catch; spoilers, nothing is free.
I hopped onto Sora on launch day and was unable to due to the tremendous interest. OpenAI halted new account creation except for folks who already had accounts. Now that it has cooled down and several days later, I hopped on to see what I could try but quickly found out that Sora, unlike other parts of OpenAI like ChatGPT and DALL-E, is only for premium subscribers. So straight off, if you want to try to create your own AI videos, it is behind a pay wall.
However, you are able to check out what OpenAI has put up into their gallery that showcases (the best) of what the platform can make as well as the prompts behind them, and get the flavor of what it might be able to do.
Access to the remix editor is possible, but again, without a paid account, you can’t really make anything.
Potential?
One of the most striking aspects of Sora is its ability to create videos based on simple text prompts. Again, keep in mind that these videos are only a few seconds long. Really, what AI is doing here, is creating several AI generated photos and learning how to string them together to make the illusion of a video.
However, Sora is not without its flaws; as AI still struggles with object permanence and physics. As MKBHD demonstrates, it often results in videos where objects disappear or reappear unexpectedly and movements that don’t quite make sense. To me, it reminds me of The Matrix, where things sometimes just present themselves just a little bit off.
One of the features of Sora I wasn’t aware of until MKBHD’s review is its ability to turn static images into dynamic videos. While it avoids generating content with recognizable public figures or copyrighted material, it can bring AI-generated images to “life”. However, with its lack of understanding of the real world, physics, and context of the source video, it makes some pretty epic (and creative) interpretations (or epic fails) of what happens in the seconds to follow.
Despite these limitations, Sora excels in generating abstract visuals, text animations, and cartoon-style characters, making it a valuable tool for creative projects, especially as all these videos that are created are without sound. But again, does it make you wonder where it gets the ideas for those cartoons?
Legality, Ethics, and Creativity
The reason why I cite MKBHD here, isn’t just because I am a long time fan of his work, but because of something he (and arguablly only he) could highlight. And that is the plant at the end. It is the very plant that showed up in his prompt to it to “create a tech reviewer sitting at a desk.”
That very plant, is in many of his own tech reviews. MKBHD has spent over a decade really defining what a “tech review” on YouTube is and looks like. Which bring up the question, where did Sora get the idea to do that?
Was MKBHD’s videos used to train Sora? Or is it an impact of his inspiration permeating the YouTube tech review space in that others mimic his style? But again, does that mean that style from others’ videos were used to train Sora? Again this simple plant, highlights the issue, and blurred line of what is “public” and what is “publicly available to use commercially.” AI companies scraping the “public” internet to train a model that they then sell to others to use, should this be legal? On the same token, this again continues to pour fuel on the fire as to the artists who created the original content in the first place that trained the AI to get that ‘inspiration’ when it is prompted: Where is the copyright line crossed, and can this be solved through regulation, legislation, or something else? Because without the training data, an AI is literally as creative as a box of rock.
As Sora and other platforms like it become more and more available to the public, it will continue to also raise other important questions about the future of AI-generated content, both legally and ethically. Let alone the fact that these tools’ potential to create convincing videos highlights the need for critical thinking and media literacy in an era where seeing is no longer believing. Literacy around the legality of the platform, as well as the ethics of its use.
As Marquez said towards the end of his video, what we see AI create today, “is the worst it will ever be.”
You must be logged in to post a comment.