iRacing is an a competitive multiplayer racing sim first and foremost, but its singleplayer AI opponents could see a significant boost as a result of a new collaboration with Microsoft Research.
The definition of AI in the gaming space is undergoing a bit of a transition. In racing simulations and games, it usually still refers to computer-controlled opponents that are on track with the player when racing offline - not the increasingly-popular and sometimes eerily human, self-learning models that can answer questions, generate images and videos, or even write stories.
iRacing is looking to leverage the latter to improve its offline opponents, however - and more: The racing simulation service has announced a collaboration with Microsoft Research "to explore this cutting-edge field". The idea for this collaboration is to not only make iRacing's AI drivers more realistic and deep, but also for driver coaches, engineers "and many other possibilities" to enhance players' experiences.
To achieve this, so-called Large Action Models are to be used as a foundation for these functions. "The goal is to provide iRacing with models that developers can use to supplement the existing product with artificial intelligence-based training, giving iRacers on-track feedback, quality in-game improvements, and new ways of improving their performance and having fun on the virtual track", as the official announcement reads.
Interestingly, the collaboration was set in motion by former IndyCar driver Oriol Servia, who will also be involved in it moving forward. iRacing President Tony Gardner states: "We couldn't be more excited to work with Microsoft on new AI-based learning models which hopefully will lead to iRacers having even a better experience, not to mention improving their racecraft and performance."
On the other hand, Microsoft research hopes to learn from iRacing to "advance our scientific knowledge and ability to build AI systems that span and connect between the physical and virtual worlds", according to Microsoft Research President Peter Lee.
The commentators used the voices of real personalities and referenced the track or drivers and their past races, and reacted to what was happening on track in real time instead of using pre-recorded lines. The race engineer used a similar approach, reminding the player of where they could improve, where they lost time the last time they drove at that specific track, and more - there is a lot of potential in building more dynamic and individual systems in this regard.
For iRacing's AI opponents, for instance, that could mean drivers with their own personalities and who remember past races agains the player - and we could imagine that they would not be afraid of retaliating for something that happened multiple races ago if the opportunity presents itself.
What would you like to see improved in sim racing based on AI, and what would be the most useful? Let us know in the comments below and discuss this news in our iRacing forum!
The definition of AI in the gaming space is undergoing a bit of a transition. In racing simulations and games, it usually still refers to computer-controlled opponents that are on track with the player when racing offline - not the increasingly-popular and sometimes eerily human, self-learning models that can answer questions, generate images and videos, or even write stories.
iRacing is looking to leverage the latter to improve its offline opponents, however - and more: The racing simulation service has announced a collaboration with Microsoft Research "to explore this cutting-edge field". The idea for this collaboration is to not only make iRacing's AI drivers more realistic and deep, but also for driver coaches, engineers "and many other possibilities" to enhance players' experiences.
To achieve this, so-called Large Action Models are to be used as a foundation for these functions. "The goal is to provide iRacing with models that developers can use to supplement the existing product with artificial intelligence-based training, giving iRacers on-track feedback, quality in-game improvements, and new ways of improving their performance and having fun on the virtual track", as the official announcement reads.
Interestingly, the collaboration was set in motion by former IndyCar driver Oriol Servia, who will also be involved in it moving forward. iRacing President Tony Gardner states: "We couldn't be more excited to work with Microsoft on new AI-based learning models which hopefully will lead to iRacers having even a better experience, not to mention improving their racecraft and performance."
On the other hand, Microsoft research hopes to learn from iRacing to "advance our scientific knowledge and ability to build AI systems that span and connect between the physical and virtual worlds", according to Microsoft Research President Peter Lee.
More Dynamic Engineers & More?
Now, all of this might sound relatively abstract, but to get an idea what the collaboration might result in, a look back to two since-deleted previews for what is now Project Motor Racing might be in order. When the project was still called GT Revival in 2023, Game Design Director Austin Ogonoski posted two video previews to his YouTube channel - one about a commentary team, the other about a race enginer, both driven by AI.The commentators used the voices of real personalities and referenced the track or drivers and their past races, and reacted to what was happening on track in real time instead of using pre-recorded lines. The race engineer used a similar approach, reminding the player of where they could improve, where they lost time the last time they drove at that specific track, and more - there is a lot of potential in building more dynamic and individual systems in this regard.
For iRacing's AI opponents, for instance, that could mean drivers with their own personalities and who remember past races agains the player - and we could imagine that they would not be afraid of retaliating for something that happened multiple races ago if the opportunity presents itself.
What would you like to see improved in sim racing based on AI, and what would be the most useful? Let us know in the comments below and discuss this news in our iRacing forum!
