Why We Might Have Digital Twins In The Years To Come

Mark Zuckerberg, CEO of Meta (Facebook), is facing his hologram avatar of the metaverse, a world built by Meta.

Atlantico: With the development of new technologies and with the Meta (Facebook) projects, the virtual universes of tomorrow could make it possible to have a digital twin. What does this technology consist of today, in its business operation? How does it work? What is it based on?

Thierry Berthier: The digital twin is the digital replica of an entity, object, physical system or biological organism. In its operation, this replica must reproduce as faithfully as possible, the mechanisms, the reactions, the interactions, the exchanges with the external environment and the transformations of the original entity. Ideally, the digital twin can be considered as the faithful projection, on cyberspace, of a source entity, without loss of information, complexity and functional simplification in its digital description. Of course, the more rudimentary the original model, the simpler the digital twin will be to produce and operate. The complexity of the source entity completely determines that of its digital replica. A human’s digital twin is, by definition, the most complex replica to produce. It must be considered as an abstract limit that we do not know how to reach with current knowledge, technologies and means of calculation. In 2022, when considering the digital twin of a human, it can only be a hyper-simplifying projection, a functional reduction, or a very incomplete digital imitation of an individual. The digital twin of a nuclear power plant, an industrial production line, an aircraft carrier or a wind turbine is the virtualized reproduction of all systems, dependency links and flows imitating, in time real, the functioning of the physical infrastructure. Concretely, all the sensors deployed in the physical entity to be imitated, produce live data which is then collected, processed and then reinjected into the digital twin. Once fed with this data, the virtual replica can be used to test hypotheses, to validate protocols, to carry out simulations or forecasts by accelerating time. The digital twin is not a simple simulation even if it can rely on simulated sequences to imitate its model. There is a difference in scale between the simulation which models a particular process and the digital twin which imitates the system identically via a double flow of data, from the sensors of the source to its twin and in the opposite direction. The interest of the twin in relation to partial simulations lies in its ability to address complex problems that go beyond the framework of the simulation. They make it possible to improve products and processes. Deployed on the data architecture of the digital twin, artificial intelligence makes it possible to animate it, to make it dynamic. The AI ​​acts as the replica’s nervous system and brain.

Read also

The great paradox of the multiverse

One of the questions that seemed like science fiction, but that some researchers are putting on the table, is whether thinking human digital twins are a possibility in the future. What about?

Generally speaking, you have to be very careful when talking about a human digital twin or duplicating avatar. Let’s start by listing what we don’t yet know how to do in 2022 (which is fiction):

First, we don’t know how to build a computer system that “thinks” like a human because we don’t understand how the human brain works or produces thought.

Second, we don’t know how to build a computer system that mimics or simulates human consciousness.

Third, we do not yet know how to build a computer system that is “aware” of its own functioning, of its states, in an autonomous way.

Since the 1960s, numerous research works, articles and theses on the construction of artificial consciousness models have made it possible to advance knowledge but have not resulted in satisfactory implementations.

This time, let’s list what is achievable in 2022: we are able to collect and aggregate all the data, metadata, voluntary or involuntary digital traces that we produce throughout our lives during our interactions with sensors and computer systems that surround us. This heterogeneous data set is our digital reflection. It was studied and conceptualized in 2017 (by T.Berthier & B.Teboul) as an overall algorithmic projection of an individual. The corpus made up of all our produced data characterizes us numerically. It evolves from birth to death of the individual and is a knowledge base for the future digital twin. When performing a query of the “First name Last name” type on a high-performance search engine (for example Google), one obtains a very incomplete overview of the overall algorithmic projection of the individual, referenced by the engine. This first set of data must be supplemented by those resulting from each of our daily interactions with connected objects, video surveillance cameras, sensors in transport, etc.

Read also

If the excesses and dangers of the real world are not enough for you, welcome to those of the Metaverse.

The global algorithmic projection is a good start to consider the digital twin without falling into science fiction. All that remains is to “animate” this projection to obtain an evolving system. We are getting there little by little with the format of avatars evolving in virtual universes (like the Metaverse of Meta or immersive simulation spaces for the training of pilots, surgeons, firefighters or fighters). The combination of the global algorithmic projection of an (inert) individual, real-time machine learning capabilities, multi-source sensors, and computational capabilities, will make it possible to animate the digital twin at a very basic level. This is a first feasible step, in the short term, towards the digital twin.

What are the possible risks or excesses, in particular ethical, of this technology if it were to see the light of day?

Before talking about the potential risks associated with the concept of digital twin, it is necessary, above all, that the virtual replica imitates, to a satisfactory level, the individual to be projected on cyberspace. We do not yet know how to approach (simulate), directly or indirectly, a model of synthetic consciousness that could “bring life” to our global algorithmic projection. On the other hand, we are going to develop new streaming learning capacities which, fed continuously by a multitude of sensors, will energize an individual’s algorithmic projection.

The risks are not all identified today. One can always imagine the neurotic disorders or schizophrenic psychoses that could occur in a fragile individual who no longer makes the difference between the evolution of his digital twin and his own existence. We can also imagine potential cyberattacks targeting the digital twin and its biological original with all the more or less toxic feedback loops that could occur during this attack. Finally, the generalized digital twin could be exploited by mass surveillance systems serving malicious actors or an authoritarian government.

Read also

Artificial intelligence: Google puts an engineer on forced leave for claiming that the group’s chatbot had its own intelligence. But why ?

In the face of potential risks, it is also necessary to take into account the contributions and benefits of the digital twin, in particular for decision support, risk mitigation, training and education or medical monitoring of the individual via his twin. The more the interactions between the digital twin and its biological model will be in both directions, the greater the benefits will be.

Conversely, what would be the possible benefits of developing digital twins, thinking or not?

To repeat: a digital twin that “thinks” like a human is science fiction today. Advances in computing, computing and machine learning will lead to increasingly high-performance, increasingly reliable, increasingly generalist AI solutions. These joint evolutions “data – AI – sensors – calculations” will accelerate the speed of evolution of the digital twin to test hypotheses, options, arbitrations that the source individual will have to make in the short or medium term. This ability to accelerate the evolution of its own digital twin opens a window on forecasting, risk reduction and optimization of decisions and their positive or negative consequences. Finally, the deployment of the twin in increasingly immersive digital spaces will contribute to virtual mobility and the cyber-physical fulfillment of the individual.


Leave a Comment