More personalised, more intuitive, more insightful: why AI is becoming the golden child of product development
Artificial Intelligence is climbing to the top of the priority list for many executives imagining the next generation of products and services. The prospect of being able to offer a more personalised, intuitive experience for customers, gather more detailed insights, and reduce the time and cost of development is understandably appealing.
But what is often underestimated is the amount of time and data that goes into making that a reality. “People want AI inside every product but they underestimate how much data AI requires,” Ying Liu, TG0’s head of AI, says. “It takes a lot of time to collect that data, and that can only happen after you’ve finalised the iteration of a hardware and/or software.”
That said, AI is elevating TG0’s technology in exciting new ways. One project, dubbed etee Pose, is using TG0’s smart yoga pressure mat technology, coupled with machine learning and AI to accurately predict a user’s pose. It’s a difficult task to do, she adds. Most other companies are trying to do pose estimation using cameras, which can recognise how someone is moving. There are privacy concerns with taking that approach, particularly with the new guidelines incorporated into the European AI Act, the majority of which will come into force in August 2026.
It’s been a lengthy project, Liu explains, but the team is starting to see some success. “We have our pressure map and we’re trying to convert that into 3D coordinates. We’ve been putting etee trackers on each joint – shoulders, elbows, wrists, knees – and then we use our own in-house library to collect the coordinates and reconstruct the human body.” It’s not something that would be possible without the processing power of AI.
The other challenge has been finding the right combination of machine learning models for the project. Because this is groundbreaking technology, TG0 has drafted in experts from King's College London and Sheffield University to assist with the development of a proof of concept. “We’re testing many models,” Liu says. One is the recurrent neural network (RNN) model, which is commonly used for sequential data processing.
Another useful tool is the vision transformer (ViT), which is designed for computer vision and is the base technology behind ChatGPT. “On top of that, we’re testing a model called liquid reservoir,” Liu says. “That helps us to reduce the size of the model, so that it can run on very limited computing power, very fast. Our hope is that the final model for pose estimation won’t need a laptop, it will be able to run on a single microcontroller embedded into the mat.”
It’s hoped significant strides will be made by 2025, and the smart pressure mat will be ready to go into production shortly after. As well as the sport and fitness community, Liu believes it will have appeal to the healthcare and gaming sectors too. “It would be great if we can integrate everything together with the etee virtual reality (VR) controller,” she says. “By stepping onto the mat, you can navigate a more realistic world in VR.”
In the meantime, TG0 is utilising AI within other projects too. There’s a knowledge transfer project (KTP) with the University of Portsmouth, using TG0’s smart insoles and AI to accurately predict ground reaction force. It will be another revolution for the health and sport science industries.
The team is also developing a customer service chatbot for its etee controller, which can offer out-of-hours support to those using the device at home. “That’s a fast project because language is the same,” Liu says about the chatbot. “We’re training that model to be able to properly answer our customers’ questions. In the future, we’ll be able to apply this to other applications too.”
“In truth, not all tasks need AI,” Liu adds. But with the right products and applications, with the necessary time, data and expert input, AI has the potential to make what had previously been deemed impossible, a reality.
