Metaverse, digital twins | Alliance

When Neal Stephenson, author of Virtual Samurai, coined the term “metaverse” in his book in 1992, he had no idea the word would spawn a billion-dollar industry 30 years later. Metaverset is a hot topic, and many retailers have already created a digital presence for tech-savvy consumers. Analysis by Dr. Biswa Sengupta, technical researcher and head of Machine Learning at Zebra Technologies.

Dr.  Biswa Sengupta, technical researcher and head of machine learning at Zebra Technologies

Dr. Biswa Sengupta, technical researcher and head of machine learning at Zebra Technologies

To put it simply, the metaverse is a virtual space with augmented and integrated virtual realities. Now it is no longer so insignificant that people, cities and countries will only exist digitally. Metaverse brings together the power of simulation technologies that we have been fighting for over the last century, building virtual worlds similar to those in video games where people can have an immersive experience.

Discover all our articles on Metaverse

The fashion world has flocked to the meta-verse. Some stores deepen their online services by allowing users to embellish their avatars, purchase products in a virtual store, and have them delivered to their homes in real life. Others have arranged a fashion week in the meta-verse, where brands present their latest creations on catwalks.

The food sector is also looking to move into the meta-verse so users can go into virtual grocery stores, fill their shopping carts, pay for and have products delivered to their homes. Also, if you feel like having lunch after your shopping, you can even order your favorite hamburger and fries.

Also read: [Chronique] Digital intelligence: some keys to understanding innovation in the metaverse era

The meta-verse becomes an additional opportunity in the world of omnichannel trading.

The meta-verse could also be a melting pot of machine learning technologies such as computer vision (VA), natural language processing (NLP) and reinforcement learning (AR). More and more companies are seeing the potential in combining voice and vision technologies for decision making. But technological advances in voice and vision alone cannot bring us closer to artificial general intelligence unless we can use these senses together to make decisions.

This means that the meta-verse can become a virtual test site, a kind of massive multiplayer game where we can build, train and implement a rich mix of machine learning technologies to develop new retail capabilities and consumption experiences.

Build it and consumers come?

Financial institutions and social media giants are among those joining retailers in the race to get their share of the metaverse cake. But more important is whether consumers want Metaverse. We must not forget that we need to spend time, money and physical energy (not virtual) on interacting with the meta-verse. At the same time, we can imagine a parallel with a video game where we create a community with our virtual neighbors, but do we have the cognitive capacity to live a parallel life? Do we not already have crowded information?

The answer is yes. Tons of companies would make money showing venture capitalists and users the promised land of the metaverse, but it would be crucial to ask what is the product? The “build it and the customers come in droves” mantra may not work. So what can work?

The first step to unlocking the metaverse would be to be able to use it to make decisions, but also to use it as a synthetic world to generate machine learning data.

You need to build the product market fit gradually, otherwise customer adoption becomes quite difficult. If industries start opening stores in virtual spaces and using digital currencies to buy / sell digital assets, users will hardly adopt the product.

Also read: [Entretien] Axa’s metaverse will better animate its employee communities

The digital twin first

The progressive step towards building a metaverse and addressing the above points is the digital twin, a subset of the metaverse. Take a small part of the natural world, e.g. a retail store, and use a simplified metaverse (the digital twin) to enable real-time visibility of all assets (goods, store employees, supply chain flows, etc.)

Then use technologies like VA to measure supply and demand in the store in real time. Automatic language processing can go through thousands of matches and tell you which tasks to perform. Finally, under the constraints of the digital twin, reinforcement learning (RL) can make decisions about how the future will evolve.

Also read: Digital twin: RTE revolutionizes the management of its industrial assets

This will allow store managers to have a real-time overview of store operations and leverage the digital twins to make action-oriented decisions. Technologically, it makes it possible to combine different vocal and visual characteristics to make optimal decisions.

The second step step to the metaverse again surrounds the digital twin, but this time for the purpose of generating synthetic data. Some technology companies and start-ups are already advocating this mindset. The central concept behind all of this is domain randomization.

A digital twin allows us to create synthetic worlds and different subsets of the same world.

For example, most deep learning-based VA algorithms require a lot of training data. The digital twin (if designed with stringency to reduce the covariance delay relative to a natural environment) can provide us with annotated synthetic data. Whether it’s millions of miles of driving data for self-driving cars or hundreds of permutations for objects under different display statistics (e.g. fruits and vegetables in the store will look different if viewed left or right, night or day). Using infographics, VA algorithms can examine all possible iterations of your fruits and vegetables. Similarly, NLP1’s algorithms struggle to generalize when the domain is random, ie. when the materials (texture, color), light direction, lighting conditions, and placement of objects change randomly. The metaverse concept could help us address some of these data efficiency issues.

In summary, step-by-step steps can lead companies to move from useful current virtual experiences to focused end-product discussions using the digital twin, so that enthusiasm for the meta-verse can build and mitigate the input delay. Exit between the product and the needs of the market. Wrong, unconfirmed assumptions can weaken companies, no matter how technically excellent the solution is.

Leave a Comment