Was the distribution of deep fake video capability to any and all inevitable? Move fast and breaks things? Better to ask forgiveness than permission? Is it okay to release a Beta Version and have the users find the bugs?
Those are not always easy questions to answer, as with just about everything the devil is in the details. In this case the details involve such things as timing, safety rails, and how quickly can society adjust to new technology expectations.
Click to access How-OpenAIs-Sora-could-change-the-internet-with-deepfakes-NPR.pdf
For example, internet development surged ahead much faster than any sort of regulatory framework could be developed.
This had good impacts as the pace of innovation and distribution was extremely rapid. It has had bad impacts in that negative consequences were often not visible until they had already permeated the internet “fabric” …and not until various type of “internet protocols and practices” had been around long enough to identify where things were going off the road.
We are now in the moment where development and distribution is no longer more speculative than reality. Sora2 Rollout has exploded into social media, and we are back wondering how well it’s all going to turn out now that the horses are out of the barn.
OpenAI’s leadership is presently taking a “we will fix Sora2” as soon as the problems (Civilizational Bugs) are discovered, and as the solutions are “tried out”. This is a scary amount of uncertainty. The timing seems questionable at best, during a period of AI pushback, and ongoing negative responses to OpenAI’s roll out of GPT5 by naysayers.
It’s possible financial realities of LLMs development and servicing would have had someone doing this sooner than later if OpenAI hadn’t done it; The article mentions Meta was close to it’s own AI video upgrades. There’s sometimes something so profoundly disruptive… that it exceeds our ability to know the problems before trying things out. Perhaps this could have been rolled out drastically slower? I’m sure that point will be debated on both sides of the question in the weeks and months ahead.
But this isn’t the last Great Leap Forward for AI that we will need to cope with and adjust to and hopefully roll out in some appropriate fashion in those same weeks and months and years ahead.