2 years

That’s my feeling too. GPT models have been around for a few years now. ChatGPT and its cousins represent one specialized use that has finally come to market. I would be shocked if the big studios have not been sinking a lot of money into this for a while.

The SAG and writers guilds certainly take it seriously.

10 years

I suspect right around the time government hops in to "regulate AI" for "safety" and other "harmful content" reasons.

You don't need the quotes.

Soon, but I don't think I, and we, will really watch them. I think we'll watch the "real" old movies instead.

I think it'll take the route of other art forms, and we will value the craftsmanship just as much as the output

The promises of the film experience at the dawn of the digital era were never realized. All of the work has been put into minimizing the costs of movie making and distribution.

Consumer display and audio technologies are the exception, but movie theaters still haven't incorporated automated ambient effects like local haptics or spatial audio, smellovision, lighting, etc, like in stage productions. There are no custom data tracks or queuing allowed in industry formats. Anything would have to be built separately and manually synced.

Despite our art and technology, there is little excess or imagination in film today, and it is desperately resistant to change.

Very cool.

Do you want to work on this? Sounds like a fun project

How long until I can get a highly accurate OCR of this PDF with open source tools?

Everyone seems to point me at tesseract, but it does things like turn:

  CHAPTER III

into something like :

  CHAPTER 1|l

Granted with sans-serif fonts, the characters themselves are ambiguous, but I have no idea how the language model lets that through.

Your assessment of open source OCR performance is the same as mine - it still sucks. Even proprietary large models like Google Gemini suck at it.

If it helps at all, someone manually did it https://www.starwarz.com/starkiller/the-empire-strikes-back-...