Now that will be peak power efficiency and a real solution for the world where all electricity and silicon are hogged by AI farms.
/s or not, you decide.
The other major success of recent years not discussed much so far is gaussian splats, which tear up the established production pipeline again.
10 years ago people were predicting VR would be everywhere, it flopped hard.
10 years ago, people were predicting that deep learning will change everything. And it did.
Why just use one example (VR) and apply it to everything? Even then, a good portion of people did not think VR would be everywhere by now.
It is odd how many people don't realize how developed self-driving taxis are.
I think most people will consider self driving tech to be a thing when it's as widespread as TVs were, 20 years after their introduction.
Fully autonomous in select defined cities owned by big corps is probably a reasonable expectation.
Fully autonomous in the hands of an owner applied to all driving conditions and working reliably is likely still a distant goal.
Those with the real vested interest don't care if that flops, while zealous worshippers to the next brand new disruptive tech are just a free vehicle to that end.
The images rendered in a game need to accurately represent a very complex world state. Do we have any examples of Transformer based models doing something in this category? Can they do it in real-time?
I could absolutely see something like rendering a simplified and stylised version and getting Transformers to fill in details. That's kind of a direct evolution from the upscaling approach described here, but end to end rendering from game state is far less obvious.
I’m really curious why this would be preferable for a AAA studio game outside of potential cost savings. Also imagine it’d come at the cost of deterministic output / consistency in visuals.
I could absolutely see something like rendering a simplified and stylised version and getting Transformers to fill in details. That's kind of a direct evolution from the upscaling approach described here, but end to end rendering from game state is far less obvious.
The main rendering would be done by the transformer.
Already in 2025, Google Veo 3 is generating pixels far more realistic than AAA games. I don't see why this wouldn't be the default rendering mode for AAA games in 2035. It's insanity to think it won't be.
Traditional rendering techniques can also easily exceed the quality of AAA games if you don't impose strict time or latency constraints on them. Wake me up when a version of Veo is generating HD frames in less than 16 milliseconds, on consumer hardware, without batching, and then we can talk about whether that inevitably much smaller model is good enough to be a competitive game renderer.
You conflate the challenge of generating realistic pixels with the challenge of generating realistic pixels that represent a highly detailed world state.
So I don't think your argument is convincing or complete.
That’s because games are "realtime", meaning with a tight frame-time budget. AI models are not (and are even running on multiple cards each costing 6 figures).
I’m so excited to be charged AAA prices for said wonderful experience.
For Genie to exceed AAA Graphics in 2035 at 60 to 120fps per second would require a breakthrough of efficiency that is at least an order of magnitude, and much higher for it to be cost effective.
The gaming industry for AAA titles requires at least 3-4 years in making. Which means AAA titles studios would need to start working on it in 2031. Possibility of All AAA games in 2031 are made with LLM model is practically zero.
We are talking 10 years from now.
And unfortunately 10 years isn't that long time in many industries. We are barely talking about 3 cycles.
It will be AI all the way down soon. The models internal world view could be multiple passes and multi layer with different strategies... In any case; safe to say more AI will be involved in more places ;)
I think it's reasonable to assume we won't see this tech replace game engines without significant further breakthroughs...
For LLMs agentic workflows ended up being a big breakthrough to make them usable. Maybe these World Models will interact with a sort of game engine directly somehow to get the required consistency. But it's not evident that you can just scale your way from "visual memory extending up to one minute ago" to 70+ hour game experiences.
The answer is clearly transformer based.
Competition works wonders.
I’d be absolutely shocked if in 10 years, all AAA games aren’t being rendered by a transformer. Google’s veo 3 is already extremely impressive. No way games will be rendered through traditional shaders in 2035.