I have talked to a Japanese professor about this. It is theoretically possible to use GPT or similar generative models to create a novel tailored to a single person and sell it to them.
Of course, some editing might be necessary. The dangerous aspect we are thinking? What if we can make an entire visual novel and lock that person in a perpetual loop of self-gratification? Pay more to unlock more chapters? Pay even more to unlock a specific route? Pay even even more to add a new love interest?
It is completely messed up, but sometimes we have this kind of Black Mirror talk as a thought experiment.
There are automated robots that can cut metal and weld, but it still requires a highly skilled person to manage them. It will take many many human lifetimes before this will be a thing you can just prompt and receive a finished product.
AI is a stupid term to use for this technology. It is ambiguous with AGI in how the public perceives it. All of the current systems are highly tuned to give reasonably good results for a few prompts in a niche area of interest. They can’t be left running for more than a few steps or the results are junk. They can’t operate outside of their niche. They require very specific prompting terminology and a detailed understanding of the software and weights to get specific detailed results. These things have complexities that are not remotely comparable to the number of neural connections in a human brain.
The real impact is as a tool to improve the efficiency of real humans. The depth and quality of film and games can really change now. Prompt engineers will become a new key professional space.
…which means you’ll have to be absurdly good at descriptions and being self-aware as to what you really want, and even then you’ll probably be both pleasantly and unpleasantly surprised.
How often does an experience, a food, a film, sound crazy good on paper but just ends up kinda meh, while something you stumbled upon is suddenly the best thing ever? We crave unique experiences; asking specifically for one doesn’t scratch that itch as well as just having one.
That’s how it is with current AIs, the more info you give them the better the outcome will be. That’s probably just gonna become par for the course. And you will probably be able to share generated content with others.
Honestly no, not without much better curation of what the next models are fed on. As is, the Internet is now getting covered with ai text, images, sound, and video. Models fed on the internet as a whole will end up consuming already ai content, which over time will result in us creating ai Hapsburgs.
It’s not that good to be honest. It felt like an excuse to bring on a bunch of guest stars and wax philosophical about reality in a way that feels as if the writers are high school sophomores.
The next two episodes are great, definitely check those out.