OpenAI shuts down its image and video generating tool Sora

[Dan Robinson]

Artificial Intelligence giant OpenAI is shutting down its Sora image and video-generating tool:

OpenAI shutting down Sora video-generating app

Disney's deal with OpenAI, which the companies struck in December, will be dissolved.

www.nbcnews.com

https://www.wsj.com/tech/ai/openai-set-to-discontinue-sora-video-platform-app-a82a9e4e

Hopefully this will become a trend. The effects of image and video generating AI systems are nearly all bad, with very little good. All do the following:

• use massive amounts of energy compared to other AI systems
• require ingesting vast amounts of copyrighted works without a license, with courts trending toward ruling this as copyright infringement
• the output cannot be copyrighted, meaning there is no viable revenue stream (no licensing, no exclusivity)
• the output replaces the artist/photographer whos works it trained on in the market (customers can generate works in that artist's/photographer's style for free)
• the output enables vast amounts of fraud and deception
• the public opinion of these types of AI gens has been increasingly negative

We've already seen many examples of systems like these ingesting massive amounts of chaser's work to generate very realistic tornado images and videos. Production companies and businesses have already reportedly used these systems in revenue projects instead of licensing from any of the chasers whos works were used to train the AI.

If the other AI giants follow suit here, this could be great news for the stock photography and video economy.

 

[Warren Faidley]

Wishful thinking but the genie is out of the bottle and AI is here to stay.

 

[JamesCaruso]

I’m glad to hear this news for a number of reasons. Knowing what’s real and what’s fake has been becoming increasingly blurred, and there’s just too much “AI slop” out there. Last thing we need is for young people to have yet another digital black-hole time-suck; Sora was even worse than TikTok in that regard.

But the problems and risks you cite Dan, around the use of others’ work, are still issues with text generation, not just video.

 

[Dan Robinson]

The more I work with AI, the less worried I've become. It really isn't one big monolith that does everything. "AI" is always spoken of as some singular consciousness (a la the Matrix), but really it's just a bunch of specialized systems that each do only one thing, and don't even talk to each other very well. For example, the coding LLMs don't do visuals very well. For workable results, you usually still have to make (or license) any visual elements of a project yourself or use one of the image gen AIs to do it (something I've never used, by the way).

I think if the image/video gen systems can be defeated with all of the existing major impracticalities and evils with them (for instance, a big movie studio wouldn't have exclusive rights to anything they generate and wouldn't have any legal standing to prevent piracy), things will look very bright for the future of photography and video.

On a related note, I've been receiving emails from licensing firms representing AI gen companies showing interest in my footage. From what I can tell at this point, these are going out to photographers and creators everywhere, and they appear to just be brokerages who will rep your work just like any other stock/ENG broker will. That is, no guaranteed sales, but they'll pitch your stuff to the AI gen firms. This doesn't appear to be a very good prospect at this point for a chaser, as it's likely to just give future stock licensing customers a cheap way to profit from your work without you. I might start a thread on it once I can learn some more definitive info.

 

[JamesCaruso]

The more I work with AI, the less worried I've become. It really isn't one big monolith that does everything. "AI" is always spoken of as some singular consciousness (a la the Matrix), but really it's just a bunch of specialized systems that each do only one thing, and don't even talk to each other very well. For example, the coding LLMs don't do visuals very well. For workable results, you usually still have to make (or license) any visual elements of a project yourself or use one of the image gen AIs to do it (something I've never used, by the way).

I agree Dan. Every doomsday scenario refers to “AI” in the abstract, as if it’s one monolithic, omniscient entity that will orchestrate everything. Most of these doomsday theories also make a lot of assumptions and leaps of faith, i.e. that AI will somehow improve itself to the point of super-intelligence. It makes no sense. A lot of this is AI companies hyping things for their own valuations.

I was listening to a Cal Newport podcast and he did some research on how people are using AI at work. It mostly came down to some useful things like summarizing/synthesizing text, writing boring administrative stuff, making slides, etc. Newport was like, this is all useful stuff, but if 20 years ago someone described this as the future state, nobody would panic about the potential disruptions. In contrast, if you told people 35 years ago that one day something called the Internet would come about, and you would no longer need to go to the store to buy something, or go to the office to work, or use a paper map to drive somewhere, and that you could communicate with anyone in the world for virtually no cost, etc. - it would have been viewed as much more disruptive and scary.

 

[Warren Faidley]

The primary players currently slowing the expansion of AI are large entertainment industries—particularly music and motion pictures—which have mobilized substantial lobbying efforts, unions, public relations campaigns, and legal resources to resist or regulate its adoption. By contrast, photographers have historically sit on the sidelines like baboons and done very little -- partially because they are grossly unorganized.

A similar dynamic was observed when Congress considered implementing a use tax on creative works, which would have applied to individual transparencies, songs, and related media. Such a measure would have created a nightmare scenario for creators. My own efforts to organize photographers in opposition were largely unsuccessful —much like the more recent response to proposed anti–storm chasing legislation in Oklahoma.