Inworld AI has raised new funding from Lightspeed Venture Partners, bringing the valuation of the maker of an AI-based character engine for games to over $500 million.

That valuation makes Inworld into a frontrunner when it comes to AI and games. The round includes additional investments from Stanford University, Samsung Next, and new and existing strategic investors such as Microsoft’s M12 fund, First Spark Ventures cofounded by Eric Schmidt, and LG Technology Ventures.

With over $100 million in total funding, Inworld AI is now the best-funded startup at the intersection of AI and gaming, said Kylan Gibbs, cofounder and chief product officer of Inworld AI, in an interview with GamesBeat. That’s a lot of money for a company built around making non-player characters (NPCs). The twist is that these NPCs promise to be as smart as people.

The company will use the funding to accelerate research and development, hire top talent, invest in infrastructure, and launch an open-source version of its character engine.

“Inworld’s commitment to open source is a testament to our belief that collaboration fuels innovation,” said Michael Ermolenko, CTO and cofounder of Inworld, in a statement. “Working with the open source developer community, we’ll push forward innovations in generative AI that elevate the entire gaming industry.”

This is all in the name of doing things that weren’t possible without today’s advanced AI. In big games like The Legend of Zelda, Gibbs hopes to turn the characters and interactions into the central part of the experience in ways that don’t happen in today’s games.

“You may have to, for example, go tell something to certain characters and learn which ones are the snitches and then change the narrative,” Gibbs said. “You may have to go and talk to a series of characters to shift their opinion on something so that they tell someone else, right in the same way.”

Characters in a living world

While AI has been around in games for a long time, the characters aren’t that smart yet.

“When you talk to a game designer, triple-A games are made to be extremely efficient. They talk about smoke and mirrors, where it just has to appear real enough for the player to believe that the world is real,” said Moritz Baier-Lentz, partner and head of games at Lightspeed Venture Partners, in an interview with GamesBeat. “To run a full simulation in the background, where they simulate this whole world, you would never do that in a game. You would just make the player believe that the character was there in the world.”

That doesn’t lend itself to much of a backstory. You wouldn’t know, for instance, that a character spend the last three days with family having a great time on vacation and now the character happens to meet you.

“Our approach is importantly different from game studios who are focused on AI with things like simulations,” Gibbs said. “We are fundamentally trying to position ourselves as the canonical toolset that anyone could use to build any game that involves some form of dynamic real world interactions. And so increasingly we’re working on this idea of a contextual mesh alongside the character engine.”

The virtual world is a living world, and the characters are portals into a storyline for understanding everything about the world. The characters can understand the world and its map, and they can reference those things in the world when they’re talking to the player.

“We’ve now gotten beyond just the personality, emotions, motivations,” Gibbs said. “Now we’ve got long-term memory. So the characters can actually not only memorize but synthesize context from very long periods of interaction. We’ve also got dialogue animations, and something that will be coming out soon is a new voice system that will be launched that is much more adaptable and realistic.”

As a game developer, you will have a no-code menu where you can manage the relationship between characters, who should talk in a conversation, and who should be the focus of attention. You can craft it so that a response is aggressive or passive or subtle.

“All of these little things add up to allowing you to have this sense of an actual living world,” Gibbs said.

The technology of the character engine

Baier-Lentz said you aren’t just going to plug a character description into ChatGPT and get all the answers you need about creating such a character. The characters carry on conversations, but they can do so because they have cognition, memory and can put things into the context of the world.

“It’s bringing characters to life in a virtual world,” he said.

Baeir-Lentz said that, with memory modules, you can specify a character’s history and characterize shared memories across a group of characters. That becomes a kind of group knowledge among characters.

Inworld AI has dialogue and character specific AI models that are optimized for games.

“And that’s important because the dialogue you have in a kid’s education context is very different from other games,” Gibbs said. “If you build it from scratch, you are replicating a lot of work.”

Creating the tech isn’t trivial. It can take customers three to four months to create models with a few engineers. Gibbs said his teams are creating templates for characters like shopkeepers that game developers can take and customize.

The Inworld Character Engine powers multimodal character expression by orchestrating multiple machine learning models designed to mimic the human brain and communication. Unlike chatbots, Inworld powers multimodal character expression by orchestrating multiple machine learning models that are designed to mimic the full-range of human communication.

The platform allows developers to link Inworld’s character brains to their animation and rigging systems, including in 3D environments. Smart NPCs can learn and adapt, navigate relationships with emotional intelligence, have memory and recall, and are capable of autonomously initiating goals, performing actions, and following their own motivations that can drive the narrative and integrate with the broader player experience.

Inworld AI trains its own models and those aren’t cheap. Companies like OpenAI have spent more than $100 million developing the large language models (LLMs) that bring intelligence to machines. A lot of the cost is in data preparation and collection, which means figuring out the kind of knowledge that you’re going to feed into the AI. There are also inferencing tasks, and all of this means spending a lot of money on graphics processing units (GPUs) and dataset preparation.

Gibbs said that his company has the benefit of being able to serve results on a large scale. So it may turn out to cost something like 0.2 cents per request, which Gibbs said is dramatically cheaper than anything else. That’s also pretty good on costs compared to a year ago. With such costs, it’s easier to figure out how to serve AI answers to millions of players, who might interact 20 minutes a day for a month with characters at the cost of maybe $1.

Gibbs doesn’t think that training custom models will become extremely cheap.

“Groups like OpenAI are working on super horizontal models, but we’re focused on very game-specific models,” Gibbs said.

These tools are created in a way that the game developers can understand them, and they’re integrated into platforms like Unity and Unreal, Baier-Lentz said.

“Ultimately, we want to be a creator tool,” Gibbs said. “We’re not trying to be a game studio. We’re not trying to be a production house. We are ultimately here so that everybody else doesn’t have to build the tech from scratch. And we don’t want to compete with them on the actual IP or everything else. We want to be that provider. We are here to work with the triple-A game studios.”

Gibbs added, “As we’ve actually gone deeper with triple-A studios, naturally, their creativity is surpassing our own. And they’re not thinking about how to take existing game designs and then improve them by just strapping on AI. They’re thinking how to fundamentally reinvent new genres of games that have never before been enabled via these types of characters.”

Among the benefits of Inworld’s character engine: You can build once and deploy the NPCs everywhere. You can create personalized interactions. You can create NPCs with no coding knowledge. The NPCs have an emotional range and in-game awareness. They have memory and recall. And they don’t give you the same answers twice.

“The next generation of games will be judged by how immersive the experiences feel,” said Ilya Gelfenbeyn, CEO and co-founder of Inworld. “The Inworld Character Engine offers a transformational shift that brings characters to life with realistic emotions and dialogue, adding richness to the stories and worlds they occupy. Populating experiences with characters that behave with convincing real-time responses and actions will play a substantial role in getting us to the promise of truly interactive entertainment.”

The Inworld Character Engine is integrated in the industry’s top game development engines and is featured in Unity’s new AI Marketplace as a Unity Verified Solution. Inworld’s developer platform introduces a layer of artificial intelligence on top of Unreal Engine’s MetaHumans to help build the most convincing characters.

Getting traction

Baier-Lentz invested in Inworld AI early when he was at his previous firm, Bitkraft. And he is investing again even though he knows there are about 300 companies at the intersection of gaming and AI that have surfaced in the last couple of years or so.

“This is the company that has the highest chance of being that generational company in gaming,” Baier-Lentz said. “And I feel very comfortable saying that in public. My job is to find that $10 billion or $50 billion company in gaming, the next Unity, the next Roblox. I couldn’t point to a single, better startup to pick.”

That’s because the company has traction not only with investors but triple-A integration partners and developers, he said.

In the video, Inworld showed it is working with Niantic’s 8th Wall, Unity, Unreal and Roblox. In addition, Inworld said it is collaborating with partners at Team Miaozi (NetEase Games), LG UPlus and Alpine Electronics. Players have also used it in community-created mods of Skyrim, Stardew Valley, and Grand Theft Auto V.

“AI-driven characters are new magic. They will power a paradigm shift (a form of renaissance) in storytelling and escapism, where the audience can transcend the role of passive viewer to active participant,” said John Gaeta, Academy Award winner and chief creative officer at Inworld, in a statement. “This type of interactive media will open up new avenues of creative expression, with narratives guided by the collective imagination of creators and the audience. Creatives will invent, spark, improvise, and guide these persistent role players, personas, relationships, scenarios, and dynamic worlds.”

The surveyors for Future of NPCs talked to over 1,000 U.S. gamers. And 99% of gamers believe advanced AI NPCs would positively impact gameplay. 78% of gamers would spend more time playing, and more importantly, 81% of gamers would be willing to pay more for a game with advanced AI NPCs.

Inworld AI was founded in 2021 by experts that have pioneered conversational AI platforms and generative models at API.AI (acquired by Google and renamed Dialogflow), Google, and DeepMind. The company has about 60 people now.

AI is the big money trend

It’s no secret that generative AI is generating a lot of investments in AI and games.

Generative AI is on track to usher in an era of tremendous opportunity and transformative change not only for game developers but for players, too, the company said. Inworld AI’s technology goes beyond existing large language models and chatbots, allowing developers to power AI-driven non-player characters, bringing depth and realism to characters and rendering them within the logic and fantasy of their worlds. AI NPCs exhibit complex and lifelike human behaviors, increasing immersion for players. Going beyond dialogue trees with conversational AI is just the start.

“In a platform shift like AI, ‘generational’ companies won’t only incrementally improve upon existing workflows with faster, better, or cheaper tools; they will create completely novel, previously impossible user experiences—like Inworld,” said Baier-Lentz, who is joining Inworld AI’s board of directors. “Stack-ranking the 200+ investable opportunities at the intersection of gaming and AI—based on upside, team caliber, product velocity, and traction—Inworld simply stands out: while everyone is circling and seeking to capitalize on the ‘new world order,’ Ilya, Kylan, Michael and the team are uniquely positioned to seize an outsized opportunity. Having previously led Inworld’s Seed round, I couldn’t be more excited to partner up with them again in my new role at Lightspeed.”

The full slate of investors to date also includes Section 32, Intel Capital, Founders Fund, the Disney Accelerator, Bitkraft Ventures, The Venture Reality Fund, Kleiner Perkins, CRV, Meta, Micron Ventures, NTT Docomo Ventures, and SK Telecom Venture Capital.

I wasn’t sure why they didn’t release the total funding in this round.

“We’re focused on the top level, completely focused on the valuation,” Gibbs said. “Increasingly, it is very clear that game studios are taking on AI as the kind of next frontier. And it is increasingly the case that while we experiment early on, we have an increasing cohort coming to us. And the story we basically have is that this character engine will sit alongside other parts of game development as a key component to the future.”

Will AI take our jobs?

I asked if the company views itself as a “net job creator” when in comes to AI and people in gaming.

“I’ve had a few conversations on this recently,” Gibbs said. “I spent last week with a bunch of writers and folks in Los Angeles, those who are writers on strike, and there is obviously, a lot of reaction against AI. When any new technology comes out, there’s the first reaction that everyone has is how can we do things that we’ve already been doing faster and cheaper. You take engines and you start replicating, taking humans that are working on a factory line, and you replicate them with conveyor belts. The first thing you do is replicate what humans already do. And then the big innovation comes when you create something that was never possible without that technology — a technological innovation.”

Gibbs added, “We are not trying to re-create games and existing forms of media with AI. We are trying to enable a new form of media that was never possible before, through interactivity, through the character designs, through building these worlds and narratives. And this is not something that has had jobs on it before. There are no people who have been out there designing AI characters and setting up these consistent world models that AI interacts with. Our hope is that we’re going to spawn a whole new class of jobs that are ultimately AI character creators, AI world designers, and that they’ll work with game developers and entertainment houses.”

Baier-Lentz said that the takeaway from screening so many comapnies on th eAI side is going back to platform shifts like the rise of the internet and the introduction of the mobile phone. Companies were born in those platform shifts that created outsized values. While companies like EA ruled earlier, there were new companies like Zynga and Rovio and King that won in mobile. In this next transition, Baier-Lentz believes the creators’ jobs will be to take a character engine and game engine build a living world.

“It will mean that creatives are more focused on actually world building and storytelling,” he said. “The complex parts of how to connect that into dialogue trees and all that kind of thing becomes easier. It means that those developers then can focus on how they set up the complex animations and action systems and everything in the world.”

A new kind of NPC

I got to experience the Character Engine in a demo at the Game Developers Conference. It was clear these non-player characters (NPCs) didn’t have canned answers to queries. I asked a series of questions and the characters answered me in conversational ways, without long delays in the interaction.

My basic mission in the sci-fi demo was to investigate an explosion and questions some witnesses. I asked basic questions about what one robot character saw and what they were doing. But I didn’t really get much out of it, and I was anxious to move on to the next witness.

Then I asked, “Is there anything else you want to tell me?” And then the robot spilled its guts and told me what it really wanted to say. It was a useful clue that I only surfaced by asking an open-ended question. That actually felt pretty fun, as it made me believe that I would get the right answer from this NPC by asking the right question, rather than just fetching a piece of information in a much more robotic way.

Gibbs said that when a narrative designer builds an experience, they can have a mental map of the sequences that a player will go through. The characters will have their internal drivers of goals and motivations. They may want to tell something to the player. The question is how they can tell that thing to the player in a believable way in the course of dialogue.

“Kids will try to break this. But as you talk to that character about anything you want, you will be told that information that you need to progress in the story and it will happen in a natural way because if the character blurts it out, it’s going to break the realism,” Gibbs said.

He noted that sometimes, in a more intricate story, a character may not tell you something until you really get to know each other. They might tell you something about their childhood only after they size you up.

“The key thing is you can think about how game design evolves,” Gibbs said. “You may get to a place in a quest where you really need to learn what the favorite action figure of a particular character is. But to do that, maybe you have to convince their sister that you’re a good person.”

Maybe you’ll have to do more quests before you can get that kind of information in order to unlock the secret.