Right now it's a tech demo at Computex, but the game-focused Project G-Assist AI may be the most interesting AI assistant I've seen yet. And I've seen a lot, and through this week's Taipei show, and through this "year of AI PCs," we'll probably see a million more.

And, by the way, you're now going to be talking about our PCs when they have Nvidia GPUs in them who are new Microsoft is ignoring Nvidia when talking Copilot+, but definitely not as the largest hardware player in the entire field. You will not be left behind by the sizzle of "AI PC". 

I'm already pretty tired of this AI story, but at the pre-show briefing, what really caught my eye was the Project G-Assist demo. It seems that the name is familiar, because it was the title of Nvidia's April Fool's "joke" in 2017, but this version is not ridiculous. This new technology demo highlights use cases that act as an AI layer on top of the game you're playing, questions about game settings and PC performance, and a

This is all the information Nvidia already has, and software overlays can be used to measure rig performance in real time.

However, the potential G-Assist layer gives you instant access to all of it through a natural language interface. Old PC hands always want to delve into the settings, tweak things themselves, run their own benchmarks, etc. And it's spectacular, but with an AI assistant that can surface all of its details for a new, more casual audience, you can instantly see

in this demo, users are asking for a graph of game delays over the past 1 minute and asking for advice on how to best set up the game for their PC. It was a great experience. It also highlighted where you can have a G-Assist that demands a minimum frame rate for the target and lowers power requirements for more efficient and cool games. It also presented potential overclocking options and resolution settings.

And I love everything. 

I'm a fan boi of PC games, and I want everyone to get as much out of the very individual nature of the hobby, but I know also that it could be a minefield of newcomers, almost exactly which you want it. 

Look at the average graphics settings screen. Often the sheer number of options that have no explanation for what they mean can be dizzying. Obviously, you have useful people like us to explain PC graphics options intelligently, but it's still a bit of research and if you just want to run your new game beautifully, now no one has time for that.

That's why I've always appreciated the GeForce experience, auto-optimizing game settings based on feedback from Nvidia and other users. If you have a Monster rig, you can bang all the graphics settings and feel self-righteous, but if you need to compromise, having a starting point to build from, or just one-touch settings to hit is very useful. I always use it on my gaming laptop and find it super convenient in those constrained environments.

So all the data is already available, and with the proposed Project G-Assist AI, you can access it without doing anything other than simply requesting it. 

Recently, my antipathy towards AI assistants has been about a lack of actual usefulness or accuracy. A large Language model (Llm) is built on a large data pool, but you choose what to present based on probability. In my experience, it made them vague at best and a complete liar at worst. But Project G-Assist, in theory, has all the clear answers to pull out to give you the right information, and is in a position to act on it for you

"think of it as something like an AI-powered version of the best settings you've found in the Nvidia app or in previous GeForce Experience," an Nvidia rep tells me. "Applying these settings on the fly in-game relies on that knowledge. Depending on the settings you are changing, you may need to restart the game, but it is not dependent on the game integration.

As a concept, I am being sold. However, when the beta is over, it's quite another thing to see it burn to version 1 of the Nvidia app. 

"AI Assistant will transform the way you interact with your favorite games and apps, and Project G-Assist will give you a glimpse of the future," said the Nvidia demo. Although for now, it is still just proof of the concept technology demo.

But there's another side to this, and there's another string in the bow of the tech demo, which is as a game-by-game assistant. Nvidia talked about the amount of guide content on the Internet for our favorite games and stated that there are 200 million guides and hundreds of thousands of wikis on Steam alone It assumes the question of: What if you could ask the developers?

In collaboration with the creators of Ark Survival Ascended, Nvidia has integrated a "knowledge database" from developers into the Project G-Assist framework. This part of the demo describes a context-aware AI assistant that can do simple things like asking where to spend skill points considering the current build or what is the best weapon in an early game.

Theoretically everything you can go to Google, but it will be presented immediately as an AI overlay during play. While I was playing, I can imagine I would have asked a lot about the Elden Ring and Baldur's Gate3, but as succinct as I did from my own search where the PC performance assistant makes sense to me as a whole machine thing, this part of the project doesn't sing so loudly to me. For a start, it must be game-by-game, since the developers need to create a dedicated "knowledge database" from scratch.

But one of the things that it can't do is crowdsourcing knowledge from real players.1 Here we are presenting only the data provided by the developer, so we don't get workarounds, tricks, or builds that the authors themselves didn't know about. This is part of the joy of a deeper understanding of today's complex games, and AI assistants, built to display only 1 source of data, are more non-useful in comparison

I can see that it is useful for basic things, but many of us rely on when we game. So our