Years? At the pace we're going, by end of the year or early mid 2024 we'll be facing the end of regular content creation and YT will be full of nothing but AI generated videos.

How is it they're going to crash and burn when they've already revolutionized industries such as programming, content creation/writing, etc? The demand for LLMs has only just begun.

I've had it argue that it can't create code, and when I repeated the same prompt to a separate instance and got code, it insisted I was mistaken.

Something has value, and it's not the product or what it produces. It's you. You're basically paying $33 a month for a bread maker. Even worse, it's not even a unique idea. The value is in getting you to pay $33 a month and only that.

https://www.youtube.com/watch?v=d8M9E_XC3So

We are already 'merged' with AI, just shitty interfaces requiring voice or text input. Neuralink and their kin will be the next step, allowing in-brain interfaces. Those will be anywhere from 'now' till 10-20 years from now. Brain Control Interfaces, depending on implementation will allow for full 'linking' and/or deep dive VR.

Because OpenAI wants to maintain control of itself, but Microsoft wants to make sure they are the next biggest owner of OpenAI so nobody else can control things.

I'd like to say between 2025 and 2035.

GPT Chat has demonstrated coding abilities. We have 3D object generation:
https://www.reddit.com/r/singularity/comments/zqnvhb/openais_pointe_prototype_text_to_3d_object/?sort=new

I had someone else say that a 5 year timeframe for AI controlled game design was 'absurd'. Modern game engines are VERY object oriented. Unity/Unreal has you write code for specific objects, and only needs to know how the object interacts (code wise) with other objects. The vast majority of single player games could be written by AI now. All that's missing is integration with Unreal/Unity, longer code output length, and higher quality assets. None of these are impossible and given the economic incentive, it's going to happen. The first company to produce something is usually the one to dominate. And with $200+ billion on the table, there's a lot of incentive.

Full Dive VR/Holodeck? No idea on when that will be. But I bet within the next 18 months we'll see the first VR 'holodeck' of AI handling coding, assets, and characters. It'll be a 5 minute paper type thing showing a techdemo/concept of someone with a VR headset having AI generate a few games from 'scratch'. Possibly like volley ball, generic racer, and generic shooter. And then progressive improvement/better, until it's a full product.

I like my language models and reddit people to not debate semantics. It can imitate to the point of letting you write a bash script that interacts with the file system, "run" the script, and the imaginary script modifies imaginary files in the imaginary file system.

Realistically, because neurons are turing complete it could be argued that it is emulation...just an impracticable and shitty one at times.

I've had it argue that it can't pretend or emulate a linux shell, opened a new session, repeated the same prompt, it worked. Went back to the original one and it just kept saying it can't emulate a linux shell...

There are forks/ports that can.