Nothing good.

As cool of an idea as the holodeck was, it just requires too much fudging to be taken seriously. Extra special holodeck matter that isn't matter (and would require a metric fuckton of energy that would put the replicators to absolute shame if it were), the ability to somehow hide the fact that the size is limited while people physically run around in it (picture quite related, that looks *way* too small) and of course the magical safety protocols (that don't matter too much in practice because they get knocked out when someone sneezes on the anyway). Yeah, yeah, don't bother linking me to the nerd stuff that's supposed to justify all of this somehow, of course you can handwave it if you want it badly enough. The one thing they surprisingly got right is the actual interactive NPC simulation itself, with characters capable of responding dynamically to the user actions. We're not *quite* there yet with LLM's (certainly no sentient characters like Moriarty), but we're not super far off either. Combine that with one of those virtual stages and some AR glasses and you can make something pretty holodeck-ish, even if the immersion would still be limited compared to actually smelling a Romulan swamp.

>>31 I always made the assumption that holodecks had built in replicator functions and the matter that disappears once reaching the arc is not gone but sent back to the replicator's current store of matter. But like you said, that's more handwaving and there's little to really make it more than just the excuse for the writers to go to other locations than the ship and random alien planets that look like California.

>>31 > but we're not super far off either Last 20% of quality requires 80% of the effort. It's basically AGI as far as I understood what you mean where there's full temporal awareness from the NPC and not just regurgitating a database with slightly more natural language. Basically, if what you're talking about was this close all chans would soon be dead after being flooded with robots that would be indistinguishable from hoo-mons. Anyway, I'd use the holodeck for vidya gaming. I'm talking Counter Strike or Battlefield type of exercise where you're on the move for an hour or two, sometimes running, practicing gunplay & RoE. Maybe (an actually good) playthrough of Oblivion with NPCs having full on interactions in a fantasy old timey world. Walking simulators would all of a sudden be a neat way to not be a lardass while doing what you enjoy.

>>34 If you have sufficient context for your LLM and some clever programming around it you can cram a surprising amount of "previously, on the adventures of Captain Proton" in there, and the natural ability of a model to hallucinate and fill in gaps on the fly does the rest. I mean for sure it's not perfect but 20 years ago I would have said it's plain completely unrealistic to have it at all, as any attempt back then to make computers sound at all creative broke down way too quickly to take seriously. I've had conversations with Richard Feynman that were, well obviously not old Dick himself, but if the way the LLM roleplayed him had been linked to a virtual representation, it might certainly keep you entertained for a little while. Enough for Barclay-style "I am the Goddess of Empathy" slop for sure. And chans are *definitely* already visited by LLM-powered bots, I've seen quite a few. Mostly clearly operated by shitposters to derail threads (we're not quite dead Internet yet), but still. The main reason we don't see bots indistinguishable from humans generating content that's not completely superficial is because there's no money in that. You want those bots shilling and honeypotting ad revenue from actual users.

>>29 That image is literally me

>>36 LLMs that can maintain context are the next thing but how long until then? Things are getting neat and weird in that space Still don't believe in singularity shenanigans though

>>57 Yeah no, the singularity stuff is people wanking over hypotheticals. In any case LLMs are *not* intelligent by any reasonable definition of the term, at best they exhibit some of the mechanisms associated with it. And simply scaling them up is already seeing limits to what "new" behavior we get out of it. When I said we were close I meant that it was close enough to provide superficial but engaging simulations that could support a holodeck, not that LLMs are already close to AGI, let alone AGI that can improve on itself. For that we'll need as yet unknown special sauce to represent the higher levels in our brains. You couldn't fire up an LLM and tell it to pretend to be Einstein and then get new physics out of it, obviously. Even its ability to handle current physics would be questionable.

There wouldn't be enough bleach in the Universe to clean the mess I'd make on every single Holodeck session.

Samurai Serials most likely and general historicals like Bashir and O'Brien did. Waterloo would be cool, maybe some Lawrence as well.

>>29 I would build my entire life there. Get a wife, a kid, grandkids even. And then as I’m dying on my death bed, I’d see a stranger looking down at me saying “your session is over”.

>>107 >he then hands you a flute and explains you were the backup plan for their civilization

>>29 I'd cast fireball