Is Nvidia’s Project G-Assist the future for gamers looking for support? © Nvidia
Nvidia is not done with its “RTX-powered” tools. After DLSS, Broadcast or RTX Video Super Resolution in particular, the manufacturer has just unveiled Project G-Assist. Nvidia’s objective, in broad terms: to allow players to optimize their games and progress smoothly using AI.
Progress in the game without leaving them
Activated with a key press or by voice via a standard sentence, this assistant which appears in the form of an in-game overlay first takes the player’s question, written or spoken. It may be a request for advice (which weapon to buy? which skill point to choose…), a request for directions on how to find your way or beat a boss, or even a question about technical performance of the game.
The operating diagram of Project G-Assist © Nvidia
Then the tool takes a screenshot of the relevant app’s screen at the time of the question. This capture, plus the player’s question and any hints provided by compatible game APIs, are sent to language models (LLMs) connected to databases (such as online game wikis).
Finally, Project G-Assist offers a personalized response, written or oral, to support the player in their request. To make it all work, Nvidia notes that the AI models can run in the cloud, or locally on a GeForce RTX PC.
Project G-Assist in Ark: Survival Ascended © Nvidia
For its Computex demo, Nvidia teamed up with the Wildcard studio to show the possibilities of its tool on ARK: Survival Ascended. In addition to those mentioned above, another example shows that the AI is able to recognize a dinosaur on the screen and give appropriate advice about it. Project G-Assist is also designed to, using simple prompts, help optimize game performance: apply the best settings automatically, identify elements that slow down a game, obtain a definition of certain settings…etc.
What deployment for Project G-Assist?
Note that this is all just an overview of Project G-Assist by Nvidia. The tool, which is probably still in the incomplete demo state, is not yet available and the manufacturer has given neither date nor details for taking advantage of it. One of the questions that arises in particular: will it be necessary to have a GPU of a certain generation to be able to have access to it? In addition, will the solution be sufficiently efficient to no longer have to leave your game in the middle of a session to look for a solution on Google? Will it be integrated into the Nvidia app or offered alongside it? Answer soon, probably.