It’s still a tool in testing mode, but Project G-Assist has already become one of Nvidia’s most eye-catching new gaming products. A chatbot with AI that can capture and analyze what appears on the screen. An assistant that you can ask anything, from which weapon is best to kill the final boss to which armor will give you more protection.
G-Assist not only provides in-game responses but goes far beyond that. During a visit to Nvidia’s offices in Munich, Germany, I tried out its new AI assistant. Although it’s a demo, it was enough to convince me that this is the natural way to integrate AI into video games.
It’s Not Cheating, but It Helps a Lot
Project G-Assist’s demo was playing the game ARK: Survival Ascended. Currently, Nvidia has only worked with Studio Wildcard to implement this tool.
A small external window on the screen allows you to ask the AI questions about the game by voice or text. This chatbot works like ChatGPT or any other assistant, but with the novelty that it analyzes what appears on the screen and has specific information about the game.
ARK is a game full of dinosaurs. So, one of the questions I can ask is, “What dinosaurs appear on the screen?” Another possibility is, “Which weapon is the best for the beginning of the game?” G-Assist answers me without a problem. It’s like having instant access to an official game wiki.
In the past, we often played games blind. But today, in almost every game, we have all the information about it available on the Internet. Project G-Assist is a natural step in this direction. Over the next few years, if you have any doubts about an object or how to pass a level, the AI will help you in a few moments.
The exciting thing is that the chatbot provides updated info adapted to each character. For example, I can ask for advice on my current equipment, what would be best to increase my skill points, and where to get more items of a particular type.
Isn’t that a bit like cheating? I asked the Nvidia team. They said no because they developed Project G-Assist with the game creators. So, if the game creators don't want the AI to tell you something, it won’t.
This means that Project G-Assist doesn’t use public data or data created by the game community, but rather the developer’s own database.
It Also Helps Players Optimize the Game
Project G-Assist isn’t just an information assistant. It also analyzes the game session, offers technical data, and improves the gaming experience. For example, I can ask it to show me a graph of the temperature data for the last five minutes or compare CPU and GPU usage.
I can also ask it to show me the average FPS rate. Here, the time varies needed for a response varied. If I ask it to do an extensive calculation, its response time will be longer.
In addition to data, I can ask it to change parameters such as maintaining a certain FPS level that puts the GPU under a specific usage, enabling DLSS, and GPU overclocking.
You can do the above because it integrates the GeForce Experience automatic optimization tool. For the most advanced gamers, this assistant is undoubtedly not enough. Still, for many others who want a quick adjustment, it’s an excellent way to adjust the game to their liking. According to Nvidia, restarting the game to make the desired adjustments may be necessary.
Project G-Assist uses different LLMs to convert voice to text. It’s currently only available in English and runs with the ElevenLabs model. It uses Nvidia’s LLM, Riva, for speech synthesis and OpenAI’s LLM for text and response generation. Still, Nvidia doesn’t specify which version of GPT it uses.
I asked Nvidia about data training. Project G-Assist can read what’s on screen, but the AI might answer incorrectly if the game changes. As with most AI assistants, G-Assist will need the developer’s help to ensure its responses are accurate and current.
G-Assist Will Be Available Next Year
Project G-Assist is a concept Nvidia says will be available to developers later this year. There’s no date yet, but it might come with the new Nvidia GeForce RTX 50, which is expected to be unveiled at CES 2025 in January.
We don’t know how G-Assist will arrive in its final version. It might come as an update to Nvidia’s app. The company says it uses little power because it only looks at the previous frame logs when it needs to respond.
Since G-Assist is an external layer, it could also work with other games. However, the tool’s charm would be lost, as the information wouldn’t be accurate enough to make it worth using—it would require the developers’ participation and the game’s data.
My doubts about Project G-Assist are similar to those about other AI tools. The potential is enormous, but it’s difficult for developers to take full advantage of it. Nvidia has created a straightforward tool for gamers, and I’d like to see it in as many games as possible. It’s a perfect example of the promise of AI. Instead of having to look things up manually, players can ask what they need to know and get the answer instantly. Whether we’re agile enough to hunt dinosaurs during the game is up to us.
This article was written by Enrique Pérez and originally published in Spanish on Xataka.
Images | Xataka On
Related | Nvidia Has Just Surpassed Apple as the World’s Second Most Valuable Company
View 0 comments