"We're Cooked"

It looks pretty convincing; it renders wrinkles, hair, and shadows well; I think it could prove useful as a tool for CGI in movies/TV/YouTube; or actors can license their likeness to be used via Google Veo, in order to increase the actor's income/output.

But I'm not sure it's appropriate to say that this is a problem. We have no idea if any AI-as-a-Service (AIaaS) will survive for long enough to cause "we're cooked"-class problems for anyone. The recession we're headed into will surely force companies to cut services that aren't profitable. Many AIaaS's are propped up more by vibes than money; they are surrounded more by fear/excitement than utility; is there any evidence that the AIaaS industry is overall monetarily profitable? Does anyone in the AIaaS industry care about basic finance right now?

I predict most, if not all, AIaaS's will be cut, especially: (a) those that are the most expensive to operate, and (b) those that appear, to most people, to be usable for harmful things (e.g. using Google Veo to generate a video of a crime in order to frame someone). AIaaS's are very expensive to operate (e.g. they require dedicated datacenters to run), and they require ever-more data to improve them. The psuedo-FOSS AIs may remain, but probably not developed further.

The AIaaS induustry has mined all data on the internet; new data hasn't been added to the internet fast enough to sustain development of AIaaS's any longer. Instead, AIaaS providers have resorted to making new kinds of ML AIs (e.g. Google's Veo); they are re-slicing the data they already have.

While you can re-slice data exponential to how much data you have, there are only so many categories of things you can generate with data: text, audio, images, video. Text-to-video AIaaS (e.g. Google Veo) is the last kind of AIaaS; text-to-video is the most difficult to get right; small flaws can show that it's AI-generated. Historians in the future will see text-to-video as the death knell of the present AIaaS industry.

What we don't know is if ML, as an AI technique, will be abandoned or not. It is normal for an AI Winter to destroy all interest in the AI techniques popular in the run-up to the given AI Winter. When you mine all available data and ML AI's still can't survive a recession, why bother with ML AI's any more? Why bother trying to use ML anymore if the AIaaS industry demonstrated ML to be a dead-end?

The Running Man was not supposed to be a documentary...