Regression Games raises $4.2M for AI gaming and esports

Regression Games has raised $4.2 million to bring the power of AI to gaming and esports competitions.

That’s pretty good for a 25-year-old solo entrepreneur with a two-month-old company. But Aaron Vontell comes from a rare breed of technical experts in a fresh area of gaming.

The Philadelphia startup is the brainchild Vontell, a graduate of the Massachusetts Institute of Technology who helped create Battlecode, an AI gaming competition for students. Battlecode was a big inspiration for the new company.

Most AI gaming is geared toward non-player characters, or NPCs, who serve as the computer-controlled cannon fodder who are slaughtered by human players. But Regression Games will let human players create AI that plays the game.

“I’ve always been passionate about AI and gaming and esports,” Vontell said in an interview with GamesBeat. “We’re creating a product and ecosystem where people can experiment, play and compete with artificial intelligence in games. And so exactly what that means is instead of using a controller or keyboard to game, players are is creating AI and writing code to actually control the characters. And the idea is to bring this really new, exciting technology to gaming and really make it accessible for people to really play around with and not only to compete but dive deeper into this exciting field of AI sports.”

Upon graduation, Vontell became a senior software engineer at NEA-backed Instabase and missed the competitive fun of applying technology to games. So he started Regression Games, which is developing a platform and ecosystem that uses AI to allow players to compete and hone their skills by writing code to control characters in their favorite games.

“I was an early employee Instabase. I worked on machine learning for low code products for extracting information from documents,” Vontell said. “So I have a background in machine learning and AI.”

New Enterprise Associates (NEA) led the round with participation from Andreessen Horowitz, BBQ Capital, Roosh Ventures and angel investors.

Vontell plans to use the new capital to grow the engineering team and accelerate the development of the initial platform to test AI-driven gaming, in partnership with existing games and game studios.

Regression Games is building the platform and ecosystem to make AI gaming and esports accessible and enjoyable for everyone. The platform will enable players to write code and AIs that control characters, debug strategies in real-time, compete for prizes in tournaments and top spots on leaderboards, and collaborate with friends to build the best bots possible.

Rather than traditional game playing where players utilize a controller or mouse and keyboard, Regression Games users will program algorithms and machine learning models to battle others. The AI platform will be made to integrate with both existing games and original games developed by Regression Games.

With more than 3.2 billion gamers around the world, 175 US collegiate esport programs, and over 47% of U.S. high schools teaching computer science, this intersection of gaming and coding is poised to grow over the years to come, the company said. By 2028, the video game industry is estimated to be worth over $435 billion, as well as an estimated 45 million coders by 2030.

“This is more like letting the consumers — the actual players — interact with these AIs. And essentially you create your own NPCs to play for you in the game. That’s a good analogy,” Vontell said. “There have been a lot of exciting AI and machine learning models coming out that really show that we can like push the limits of how we compete in games and how we play in games.”

A good example is the OpenAI 5 machine learning model that was developed to play Dota 2 in 2017. It showed that AI could become really good at playing games.

Vontell envisions varying levels of player expertise. One competition could be for people who are new to coding, while another might be for machine learning experts on the high school or collegiate level. Vontell is hiring.

Origins

Vontell has spent the past few years working on low-code / no-code systems for non-technical people to harness the power of large machine learning models.

He started studying AI in college as early as 2014. He took an interest in startups and he started playing Battlecode as a freshman and eventually helped run the club. Battlecode was an annual AI gaming competition at MIT where college students built virtual bots to compete in video games.

“People really enjoy being able to use their technical skills to compete and win prizes, and with the proliferation of AI–this is the perfect time to start bringing AI gaming to the masses,” said Vontell. “There is a lot of really great work coming from the ML community in using neural networks to play games like Minecraft, Dota 2, and Atari games. I want to bring those ideas to the mainstream and give people an accessible place to have fun using and competing with artificial intelligence in games.”

It’s unusual for VCs to fund a company is such an early stage, but NEA’s Rick Yang, general partner and head of consumer investing, said in an email to GamesBeat but they were impressed with his work at Instabase, where NEA was also an early investor.

“We’ve funded a few founders coming out of Instabase, which prides itself in bringing on very entrepreneurial talent. His technical chops and talent were apparent at Instabase so we had conviction in him as a founder,” said Yang. “We’ve been thinking about the area that Regression is building in. I love the idea of a metagame that is about ‘hacking’ the game, and there’s a high overlap of coders and gamers. We also validated a lot of the ideas around bringing hardcore AI to games when talking to several trusted folks at large game developers.”

That made it a fairly easy decision, Yang said.

Vontell believes that the tools and approaches for AI competition are useful to more than just the players, but also to game studios that wish to enhance AI capabilities, or run their own AI competitions.

“A lot more people are gaming these days, and there is a lot of complexity that goes into planning the NPCs and AIs that are used in the metaverse, online battle arenas, etc,” Vontell said. “The tools and platform we’re building for players to compete will also be extremely useful for developers who want to build AIs into their games. It’s widely known that existing bots in games are lacking, and we can do so much more than what is currently implemented.”

Regression Games plans to hold a private alpha test and tournament with players on its AI platform in the coming months. Interested players and game developers can visit here to sign up for the mailing list.

“I’m excited to get this out and see what the players think,” Vontell said.

Sapient AIs?

Does Vontell believe sapient, or human-like AI is on the horizon?

“It’s an interesting question and I discuss this a lot with my friends who are in this space,” he said. “Everyone always has a very different answer, like it’s already here or it’s 100 years away. I don’t have a concrete answer, but I think we have to define the goals we are trying to hit. What’s our definition of sentience? You can make an AI that can play a very simple Pong game, and it is going to be very different from League of Legends, right?”

AI has a tendency to scare people, given dystopian science fiction. And Vontell said that people can find malicious ways to apply technology. But he believes regulations can protect us as we build out AI technology.

“At Regression Games, we make sure that we’re building up these AIs in a responsible way,” he said. “But that’s an important question to be asking ourselves.”

Vontell believes that gaming hasn’t fully exploited AI. He wants to take it and put it in the hands of players with a new twist on user-generated content.

“There is a wide spectrum of ways where AI can be applied to games and so the way I think about it is that AI is very modular,” Vontell said. “And the way that you can apply it, imagine mixing and matching these different approaches for different pieces.”

For the roadmap, the company will expand and create some application programming interfaces (APIs) to get AI players working in existing games. That will let players start competing and building the community, and it will generate feedback from players.

“Then we’ll have conversations with studios and game developers and players to develop the product further and get it in the hands of more players and game studios,” Vontell said.