The game changer for AI avatars - Digital humans: Out of the box!

Pearl is a metahuman, which is currently the highest level of a development that started years ago with relatively simple activities. “At that time, among other things, we used a drone and AI support to determine the area affected by toxic giant hogweed in Denmark,” recalls Nørmark, now Global Head of Innovation at NTT DATA Business Solutions. This was later followed by various tools including a robotic gardening system led FarmBot, communication AIs for a child protection association and for people with dementia, a home school helper for the coronavirus period, and an AI nose for industry.

“Welcome, how can I help you?”

Alongside these tools, Thomas Nørmark’s team continued to develop the innovation department’s flagship concept: the digital human. This avatar was and is being used in countless versions, from the front desk in a Danish car dealership (“Kia Mia”) to government authorities and works canteens (“Aiko”) to a digital guide on stages of the Tour de France (“Marianne”). Always with the aim of welcoming people personally and providing them with the information they seek. For a number of years now, it has also been equipped with a GPT language model. “Today, more than 20 of these avatars are in operation worldwide,” says Nørmark. Pearl, who worked at the Basketball World Cup, belongs to the next generation of digital humans whose appearance differs significantly from that of their predecessors. “For over a year, we’ve been able to develop high-fidelity avatars much faster and at a significantly lower cost,” says Nørmark. In the past, you had to go to a photo studio, but nowadays smartphones can be used as 3D scanners. “Among other things, that’s how we created an avatar of our Group CEO Kaz Nishihata.” The scan is placed as a kind of mesh over prefabricated characters and the digital human is ready.

A game-changing platform for metahumans

The 3D tool Unreal Engine, a leading platform for video games, made it possible. The manufacturer developed the MetaHuman framework that people and businesses can use to create their own avatars. “It was an absolute game changer,” recalls Nørmark, “and an optimum concept for us because we can use everything on one platform out of the box.” Visually, everything is possible: You just need an idea of appearance, suitable images, and a little fine-tuning to turn Pearl into her sibling Edo – also a metahuman, but very different from Pearl. He works in a job center and looks professional, not sporty. Nørmark says: “Be it hair, skin color, looks, or freckles – external appearance can be adjusted quickly and efficiently depending on the application scenario.

The avatars’ intelligence is based on the concept of AI agents, which Nørmark considers the best strategy for optimal output. AI agents can autonomously complete tasks, process data, make decisions, learn from the results, adapt, and interact with their environment, all without programmed rules. The agent selects the necessary resources based on the context and user requirements, says Nørmark. “Because it knows which tools are most suitable for solving a task.” For example, to calculate the distance from a hotel to the train station, a location tool such as a Google API is required. The result is then provided using a large language model (LLM) such as ChatGPT-4o. If further information on sustainability or guidance on pets is needed, the AI agent generates an SQL command to search the hotel’s database. “There was also a game changer in these areas in 2023.”

From research to delivery

As development progressed, the innovation team at NTT DATA Business Solutions also changed significantly, says Nørmark. “We’re switching from a technically oriented approach for research and development to scaling and to the roll-out of avatars.” This requires a large, professional product team and global delivery – in Europe plus in the Philippines and Japan, as well as in the USA in the future. One reason is, says Nørmark, that the GenAI revolution has made customers much more open to avatars and the technology as such. Demand for their integration into websites is growing “as AI avatars are much better than annoying chat-bots”.

AI market is moving forward

Meanwhile, Pearl has found a new job at the New Zealand Campus of Innovation and Sport. “We are now in a very good position to shape AI development and the competition from the front,” says Nørmark. Four crucial factors converged at NTT DATA Business Solutions, he says: an early start, financial resources, technical expertise, and the global strength required to implement projects quickly and efficiently. In addition to the metahuman core team, there are currently an extended team for delivery in Denmark and centers of excellence in Manila and Tokyo, plus a number of experts who can provide support worldwide – around 50 AI specialists in total. The goal is far from being achieved – after all, every new technology creates more challenges. However, Nørmark is in no doubt that metahumans are the future of interaction between people and machines. He is now working on the features of the next generation: “Imagine the person being reflected in the eyes of the avatar in real time – that’s how a deep emotional connection can be formed.” It is a tough challenge that requires a lot of computing power, admits Nørmark. “However, given technological progress, we should be able to manage it within a few years.”

Learn more about out digital human platform PARSONII