[idiot_predictor]

* One or a few of the faangmanga-ish leaders will retire (or forced to by the board) and the new leader will be from within (likely Tim Cook for age reasons but could equally likely be Sundar Pichai for non-age reasons)

* LLMs (for non-coding tasks) will likely fizzle out as expensive talking fidget spinners and not the world saviors that the companies behind them envision them to be. AI will go back to being fun and exciting again and not the delight of both Wall Street and the ‘shoeshine boy’.

* Cloud egress costs will be heavily scrutinized and competed upon as transferring data no longer becomes a competitive advantage.

* Apple or someone like Apple will take advantage of cratering storage/compute costs by moving things off cloud to a locally owned ‘box’ that pairs with a new dumb-client/thin phone (backups, massive storage, running some compute on this box, sharing/storing ‘family’ stuff, streaming games or movies etc) also accessible via a reverse proxy (vpn) from anywhere.

* A combo of aibo/roomba/ring like device that goes about or rolls around your house and does things for you… kinda real life flubber.

* Selling users’ data goes out of fashion (goog/meta) as new companies use privacy as a competitive advantage to sell ads.

* Finally games actually have AI that doesn’t suck - a new class of AI games that would challenge and delight humans like never before.

[tasuki]

> LLMs (for non-coding tasks) will likely fizzle out as expensive talking fidget spinners and not the world saviors that the companies behind them envision them to be.

100% wrong. LLMs are wildly useful in daily life, even the cheaper models.

[mattkenefick]

For real. It's always a headscratcher when I see these weird takes against it.

Amazon has been working with Anthropic which at the very least will create a much more comprehensive Alexa.

> expensive talking fidget spinners

Like Alexa, Siri, and Hey Google? Which are used A LOT and will only be used more when they're natural, smarter, and conversational?

Always a strange take.

[idiot_predictor]

Alexa Google assistant and Siri were all significantly down sized in the last year or so (teams, product).

Turns out the top queries billions of dollars went into were to set timers or to play a song.

I would love to be proven wrong though.

[emptiestplace]

I think part of the problem with these assistants is they are fucking annoying to use for all but the most trivial tasks. The only thing I've ever used Siri for more than once is "Siri, two minutes" - and sometimes she still tells me I don't have the timer app installed. The moment I can reliably get an answer to something I'm contemplating, it's going to have a huge impact on how much I use my phone.

[anastasiaess]

I think that has to do with the nuance of programming a voice based flow that works for everyone. LLMs are on a different UX bandwidth.

[disambiguation]

+1 for AI games.

I'll add to it that we'll see more gamification of AI training as well. I.e your play data becomes training data.

[jasondigitized]

Infinite side quests with AI are going to get wild.

[monkeydreams]

I can't see how, not unless it is a 'AAA' always online game or if the quests are cached.

I think the most amazing use of AI in games (specifically RPGs) will be with statistical models to crunch "planning" and "cause and effect" to draw players into plots, schemes and storylines and then to demonstrate the impacts of success or failure of the player's quest. Think economic simulations of a city due to the player's actions or sabotage where the player has crippled a port leading to the NPCs barking about hunger, prices and disease.

[ThatCreeper]

1. I don't know

2. I think this is probably right-ish. From what I hear LLMs are nearing what they do and I have never found any interesting use other than just laughing at them (which I could do with markov chains). AI-generated images are hideous and I have no opinion on AI-generated music as I just choose not to expose myself to it

3. Again, don't know

4. God I hope so. That would be such a funny trend. Plan9 comeback??????? #define nil (void*)0 in common parlance???

5. I feel like this is exclusively a novelty but what do I know

6. No it will not???

7. This will probably happen, but it won't be many games. It will not be a new class of games. AI--big shock--isn't fun. Mario but the levels are incomprehensible. BALATRO but I CONVINCE the GAME to give me INFINITE MULTIPLIER? (Part 3). The Beginners Guide but no one cared enough to tell a story. AI so far has been a hindrance where high quality work is concerned. Why will that change? The paint bucket tool has been slowly improved over time, but it was helpful when it first came out.

How many people do you know that have played the AI-generated Minecraft thing more than once or twice? It's interesting but utterly bland and bizarre. You're constantly having to tell the program over and over again what it means to be a person when you're in a world which doesn't even know what bird is in the sky. It gets infuriating.

The idea comes back again and again, though. "What if we made a game with infinite content so it could appeal to everyone? And then what if we added multiplayer?" I've thought it, you've thought it--everyone has. The important part is realizing that you're wrong. If everyone's had this thought, why don't we have this game? Well, we do.

Dwarf Fortress is perhaps the best example of a reality simulator standing at simulating a whopping 42% of everything ever. But if you asked someone to describe Dwarf Fortress they wouldn't just say "it's a pretty good reality simulator"--well, they might. Assume here you asked them to give a /good/ description--they would tell you it's a tough, slow game about minion dwarves going around and doing your bidding in order to build a civilization which will eventually fall. And they'd tell you that it's fun because of its deep interactions. But now we've lost the plot. It's not a game for everyone anymore, because now it's only for people who like slow games with deep interactions. And it's not a game where you can do anything because you can only do so much as your dwarves can do. But Dwarf Fortress is a fun game. And the fun of it comes from its rules. And rules are what AI is famously bad at.

Minecraft is another famous game and the same thing applies to it. "You can only interact with what your player can see," "one block at a time," and "everything is a full block" are all rules which bend it away from that "ideal" game, but they're all the things that make Minecraft fun. They're all the rules.

Games aren't fun when you can say "well I have better armor" and get out of a tricky situation. Games are REALLY FUN when you can say "wait, I have a bag of holding, don't I?" and then look around as everyone starts smirking at the implication.

That's what AI is bad at. There will not be a new class of AI games.

[idiot_predictor]

6. Everything else is a prediction but 6 is my secret wish for Santa. Not gonna happen is it? :)

7. I would love to see a world that develops into something like dwarf fortress that uses AI instead of hard coded rules. And make it accessible too. By AI I really mean something like RL not the hyped up LLM craze.