OpenAI has released some of the most popular AI models we have seen so far. Joining ChatGPT and Dall-E in the company’s suite of generative AI is Sora, “an AI model that can create realistic and imaginative scenes from text instructions.” OpenAI’s model improves on the shortcomings of other models like Runway AI which launched in April 2023. The arrival of an AI model with capabilities as vast as that of Sora, brings in a ton of questions that will need to be answered, especially around the safety of this model. So, let’s have a look at some of the videos that demonstrate the capabilities of this model, and how OpenAI is making an active effort to combat misuse of a model like this.
After announcing the launch of the model, OpenAI published a bunch of videos on their website, https://openai.com/sora, demonstrating the capabilities of OpenAI’s Sora. Sam Altman, the CEO of OpenAI, after announcing the release of the model, also put out a Tweet asking people to send in prompts to generate videos, and the videos that were generated took me and others on the internet, back by surprise. Have a look at some of the videos posted by Altman on his Twitter timeline generated using Sora –
How to get access to Sora and how safe is it?
The first thing that would come to your mind after looking at these Hollywood-esque videos would be the myriad of safety concerns that surround a model like Sora. We have seen generative AI models in the past being used in all sorts of ways. However, with a model as capable as this, OpenAI has ensured that they are taking all the precautions necessary from the get-go.
First off, as it is with any AI model, access to Sora is limited at the time of launch. The company also mentioned in its release blog that they are actively working to conduct tests that will focus on ensuring that the most critical scenarios around this product are put to test. They said, “We are working with red teamers — domain experts in areas like misinformation, hateful content, and bias — who will be adversarially testing the model.” The videos are also limited to 1-minute and marked as – generated by AI.
The checks that are in place, will be done at multiple levels by OpenAI. In addition to the checks done on the prompts they will also be doing checks on every frame of the video that has been generated using Sora. They added, “We’ve also developed robust image classifiers that are used to review the frames of every video generated to help ensure that it adheres to our usage policies before it’s shown to the user.”
With the release of Sora, OpenAI has the experience of dealing with adversaries trying to game the model, having faced the brunt of the misuse of its earlier models like ChatGPT and Dall-E. All these models have become smarter in rejecting queries that push their boundaries of safety and security concerns. Not only that, they have also gotten better at what they do. So, the road ahead for Sora also looks the same, with the only difference being that this time, the bumps along the seem that they’d be avoided better than they were in the past. All we can do is wait and hope for the best.
Follow Us
Satvik Pandey
Satvik Pandey, is a self-professed Steve Jobs (not Apple) fanboy, a science & tech writer, and a sports addict. At Digit, he works as a Deputy Features Editor, and manages the daily functioning of the magazine. He also reviews audio-products (speakers, headphones, soundbars, etc.), smartwatches, projectors, and everything else that he can get his hands on. A media and communications graduate, Satvik is also an avid shutterbug, and when he's not working or gaming, he can be found fiddling with any camera he can get his hands on and helping produce videos – which means he spends an awful amount of time in our studio. His game of choice is Counter-Strike, and he's still attempting to turn pro. He can talk your ear off about the game, and we'd strongly advise you to steer clear of the topic unless you too are a CS junkie. View Full Profile
