Gamers have been clamoring for better non-player characters (NPCs) for years, and the arrival of conversational AI may finally provide the computing superpower to make it possible. Several companies are now using natural language processing AI for games and entertainment, customer service, training, and education. Mindverse’s MindOS targets enterprise, while Inworld helps game designers create AI-powered NPCs (nonplayer characters), which they describe as “Mind as a Service” (MaaS).
I saw Inworld’s extraordinary AI at work while attending the Disney Accelerator demo last fall where Inworld powered a very, very chatty and diplomatic 3CPO robot.
Inworld AI was founded two years ago by Ilya Gelfenbeyn, Michael Ermolenko, and Kylan Gibbs, who met while working at Google. Their Chief Creative Officer is John Gaeta, a former Magic Leap SVP who most famously won an Academy Award for his SFX work on The Matrix. Last year, Inworld introduced a demo game called “Inworld Origins” in which players take on the role of a detective in a dark future city, where humans mingle uncomfortably with robots.
In a recent interview with Forbes, Gaeta and Gibbs talked about their use of AI and semantic understanding to create human-like interactions with virtual characters, emphasizing the importance of personality, context, and relationships. “Writers will create personalities, and the personalities will create the response to the player. They’ve got a goal, not a script.”
Inworld also has a gallery, Inworld Arcade, where Creators can share their work with everyone. “Empowering creative developers is our core mission,” said Gibbs, who pointed to the Inworld arcade, a public showcase of experiences and characters that people have created. Recent additions to the Arcade include a Skyrim mod, an Action RPG, and a brand mascot. Their business model is usage-based.
Singapore-based generative AI company Mindverse came recently out of stealth and launched the closed beta of MindOS, a tool for creating embodied AI, or virtual beings, which targets business applications like sales and service. The characters MindOS generates are customizable and easy to train on company products and knowledge. Importantly, its designers are able to train the application by inputting disparate, unstructured data like manuals, business documents, books, websites, spreadsheets and other documents. This dramatically reduces the time it takes to spin up a new AI mind to power a 3D avatar.
“At Mindverse, we are all about making life better with AI beings,” said co-founder and CEO Dr. Felix Tao. “We are building MindOS with empathy so they can really understand users and their needs, boosting productivity with AI copilots so people can focus on being creative and collaborative.”
“MindOS AI beings are more easily trained in industry-specific knowledge and firm-specific policies,” said Mindverse Co-Founder and COO Kisson Lin. “The core technology of MindOS is fine-tuned, controllable LLM that allows our users to balance conversational fluency with fidelity to a brand’s voice. We see our role as acting as a bridge between unwieldy LLMs and the very specific needs of particular industries, hosted in an AI-native environment that allows AI agents to collaborate with people and other AI agents.”
“Nothing will compare to the immersion potential unlocked by natural language processing,” said Tech Entrepreneur Christopher Travers, who previously founded Virtual Humans, which is focused on avatars as influencers, some of which are powered by AI. “By giving NPCs autonomy and intelligence within a well-written context, it will be like adding a new, key element to the periodic table of the gaming world—all past mechanisms will be elevated as we also unlock entirely new combinations of immersive experiences.”
Earlier efforts to create conversational AI go back several decades, as this 1987 digital assistant concept video from Apple illustrates. Twenty years later Apple introduced Siri, the digital assistant everyone loves to hate. It still can’t do most of the things promoted in its splashy 2011 launch commercials starring offbeat celebrities like director Martin Scorcese.
Samsung Labs spun off Neon in late 2019. The new company made a splashy debut at CES in 2020 with virtual humans so nuanced and lifelike they seem to be volumetric captures of actors. They are not. Today there’s nothing recent about Neon online. I did find Neon’s former CEO Pranav Mistry on Linkedin, where his bio now says the former Samsung executive, working on a start-up. Neither Mistry nor Samsung replied to email requests for comment.
Former Oculus Story Studio executive turned producer, Ed Saatchi’s Fable studio, produced a VR film experience in 2019, Wolves in the Wall, based on the graphic novel by Neal Gaiman. It starred an AI character named Lucy. In VR you played her imaginary friend. After giving the Lucy AI a life outside the VR experience, texting on Facebook Messenger and talking on Twitch as if she were a real pre-teen actress. Saatchi turned his attention to building a simulation for Lucy and other Virtual Beings to live in. Unlike other chatbot creators, Saatchi believes that “AI people” should not just be text bots, but have rich, daily 3D lives in simulations, even when we are not using them. The project, called “Lucy AI,” was just nominated for a 2023 Peabody.
Charlie Fink is the author of the AR-enabled books “Metaverse,” (2017) and “Convergence” (2019). In the early 90s, Fink was EVP & COO of VR pioneer Virtual World Entertainment. He teaches at Chapman University in Orange, CA.