NVIDIA’s ACE AI has recently been drawing everyone’s attention at CES 2025. The Las-Vegas-based Consumer Electronics Show (CES) is frequently the perfect location for tech companies to show off their latest and greatest innovations to the world. These innovations range from electric vehicle concepts like Honda’s futuristic 0 Series to a $1,700 robot vacuum (yes, really). With such a broad showcase, it is inevitable that video games would have a part to play in this celebration of technology.
NVIDIA, being one of the premier producers of GPU technologies, has expectedly leaned in on the AI revolution that has been taking the tech industry by storm. They’ve begun their journey in AI with their Deep Learning Super Sampling (DLSS) image enhancement and upscaling technology. Since then, they have been infamously utilizing their updated versions of DLSS for further upscaling and frame generation to simulate high frame rates and high resolutions on hardware that could not previously handle such tasks. DLSS technology has been criticized for enabling AAA developers to disregard optimization as DLSS and other similar technologies will fill the gaps that poor optimization leaves by literally generating the desired frames and pixel density.
Regardless of the criticisms and problems that DLSS has brought to NVIDIA, they have decided to double down on AI in an interesting way. Their reveal of NVIDIA ACE at CES 2025 has been a major topic of discussion in the video game world.
NVIDIA describes the ACE AI suite on their website and how it can be utilized in the creation of video games and how gamers will interact with the virtual world. In short, ACE’s small language models (SLMs) are less hardware intensive and less complex than the large language models (LLMs) that are more commonly used by ChatGPT, Google Gemini, and Microsoft Copilot. The reasoning for this is so that the AI’s language model can carry out processes on the card itself, rather than sending it to a server where it will be processed and sent back over an Internet connection. Additionally, the ACE suite has been specifically designed to carry out these video-game-specific processes in real time. These processes are supposed to mimic the same way that humans think in three steps: perception, cognition, and action. I’m keeping it really simple and you can read more about how the tech behind NVIDIA’s ACE works here.
More interestingly, and within my realm of understanding, is the real time implementation of AI in some upcoming video games. NVIDIA has announced partnerships with several game studios and revealed what they have been working on using ACE AI.
The upgrade to traditional AI companions through the use of SLMs is shown off in PUBG Ally. When in the past, CPU allies were usually in the way more than they actually helped, PUBG Ally claims to have drastically improved the logic and interactivity players have with their in-game partner.
Labeled “the first co-playable character,” the ally can call out enemies, fetch resources, and seemingly engage in dialogue with players. While impressive at first glance, it has yet to be seen how deep the logic behind the PUBG Ally really goes. Is it possible to develop strategies or train your ally to adopt a certain playstyle? Then balance questions arise. If the Ally is intended for use in PvPvE play, it is essential to find a good balance of usefulness and the possibility of the AIs unfairly carrying players to a win.
The AI companion talk doesn’t end with PUBG, as NetEase’s upcoming Naraka: Bladepoint has implemented an AI companion system of their own. Naraka: Bladepoint being a battle royale game raises similar questions about balance, too, and, similar to PUGB Ally, this companion can search and fetch weapons and resources. I found something interesting taken directly from NVIDIA’s article speaking about “making plays that’ll help you achieve victory.” This implies that there will be a level of strategic development with your AI partner. The ability to train an AI partner and strategize will probably be the most significant innovation brought on by the use of generative AI in video gaming. Nonetheless, NVIDIA shared many other ways that AI is changing the industry.
RELATED: AI VTuber Neurosama Beats Twitch’s Longest Hype Train Record
MIR5 by WeMade has a unique application of AI in their boss enemy, Asterion. This monstrous enemy is said to learn and adapt its battle strategy based on the player’s approach to the battle. It is supposed to be capable of identifying a player’s role in the party and respond accordingly. For example, if a player has taken the role of a healer, Asterion may single this player out and try to eliminate them before any other players. It would be interesting to interact with an enemy like Asterion and see how complex its AI really is.
With any luck, AI-powered bosses may become the new norm, but it is just as likely that there may be an exploit to combat enemies powered by AI. Only time will tell if enemies and allies will truly benefit from AI. But one thing will certainly gain a layer of depth, and that is simulation games.
Krafton’s sandbox life simulator InZoi has announced Smart Zoi co-playable characters. The Smart Zoi are capable of carrying out a multitude of tasks and interactions with other NPCs. Then, at the end of the day, they ruminate on the day’s events and plan for the next day accordingly.
This shows a level of “thought” for NPCs that was unheard of before generative AI, considering the amount of writing, voice work, and programming that it takes for a single NPC in games like The Elder Scrolls V: Skyrim. Now with ACE, it is possible to produce an entirely unique character with a fraction of the work.
Speaking of cutting costs, AI art has had a pretty bad reputation since its inception. Seen as soulless and unethical by many, AI image generation has been immensely popular regardless. It is no question that studios have been experimenting with AI in an effort to provide gamers with as much content as possible for the lowest cost. In the case of TiGames’ ZOOPUNK, they have implemented methods of customization with the assistance of generative AI. Instead of having developers create assets and textures, TiGames has built a system where a player can communicate their desired spaceship paint job to an in-game AI. And it will use AI technology to generate and apply said paint job to the player’s ship. There is even an option to describe a sticker or logo to the AI and have it generate and apply it on the ship’s side.
The concerns for this utilization of AI in ZOOPUNK are obvious. Even though it is a small part of the game, one has to wonder how far a studio will go with AI when cost of labor is an ever-growing concern. Consider the last example I have from NVIDIA’s showcase. Developers showed off their ability to use ACE tech to power NVIDIA’s Audio2Face technology.
This technology streamlines the facial animation process through the use of AI by inputting an actor’s voice lines and generating animations for a character’s entire face. This brings me to another interesting quote from NVIDIA’s article, which mentioned “saving developers considerable time and money.” This can be seen as a red flag about the automation of jobs in the tech industry.
Perfect World Games’ MMORPG, World Of Jade Dynasty, uses this tech in their facial animations. And frankly, it doesn’t look terrible. As we have learned from the AAA gaming industry, serviceable is good enough to ship. It would not surprise anyone if one day we learn that design, programming, and animation will all be automated in the near future.
I genuinely see some awesome innovations in terms of NPC interactivity. Though not to leave on a sour note, I find it difficult to be excited by the possibility of swaths of intelligent NPCs in an open world when the possibility of cutting countless jobs in the industry is just as feasible. It should be noted that the majority of the studios featured in NVIDIA’s showcase are from overseas. It is very likely that there aren’t protections in place for those whose jobs are subject to automation. Either way, it is not likely that laws will completely stop the progress of AI in creative spaces but rather only hinder it. In order to keep the industry employed, we can only hope that there is a demand for “hand-crafted” video games in our future.