Latest technical reports: Apple enters the smart home market, the limits of digital twins, and a new vital communication system

The American electronics and technology company “Apple” intends to enter the smart home devices market on October 13, after a long delay, which represents a vital expansion of the company’s product portfolio under the leadership of its new CEO, John Ternos, according to what was reported by the German News Agency.

According to informed sources, the main device in the smart home package strategy has the code name “J490.” Apple also plans to announce the first update to the HomePod Mini device since its launch in 2020, and the first new version of the digital TV receiver since 2022.

Bloomberg reported that this represents an early and important test for Ternos, who took office on September 1.

The company views the field of smart home equipment as a profitable opportunity and one of Apple's famous “next big thing” initiatives. At the same time, Apple had previously faced difficulty in achieving a strong presence in this sector, lagging behind Amazon and Google.

These products are also a showcase for the capabilities of the new Siri digital assistant powered by artificial intelligence, a technology that the company has been developing for years. The launch of the updated version of Siri has faced numerous delays, and smart home devices would highlight Apple's efforts to finally catch up in the field of artificial intelligence.

According to sources - who requested anonymity, since the products have not been officially announced - the Home Hub central control device will take the form of a square screen measuring about 6 inches, with versions available that can be installed on the wall or placed on flat surfaces.

At the same time, the new versions of the HomePod Mini and Apple TV will retain their current designs, but will contain faster processors to support stronger artificial intelligence models in Siri.

Can artificial intelligence know a human being to the point of predicting what he will think or choose in a situation he has never encountered before? Researchers from Columbia University tried to test this idea by creating “digital twins” of real people, but the results revealed that the road to digitally replicating human behavior is still long.

500 questions to build the “twin”

The study, published in the journal Science Advances, included about 1,800 people, each of whom answered more than 500 questions covering personality, cognitive abilities, economic preferences, and biases, in addition to demographic and social information, according to what was reported by the French newspaper Le Figaro.

The researchers used this data to feed a large linguistic model with the aim of creating a “digital twin” for each participant. After that, the real people and their digital copies were subjected to 19 new sets of tests, which included 164 answers and behaviors, with the aim of knowing the ability of artificial intelligence to predict their actions in new situations.

Hundreds of pieces of information did not improve the accuracy of the simulation

At first glance, the results looked encouraging, with the model that had the complete data scoring 0.748 on a scale that compared its answers to those of humans.

But the surprise was that a model that only obtained 14 basic demographic information about each person scored 0.746, while a model that did not receive any personal information achieved 0.734.

Thus, the study showed that providing the artificial intelligence with hundreds of pieces of information about a person did not make it significantly more able to imitate his behavior, compared to a model that received only basic information about him. The comparison also showed that the ability of the “digital twin” to simulate individual differences between people remained weak.

More stereotypical versions of humans

The study revealed that artificial intelligence tends to “smooth out” differences between people. Out of 164 indicators, digital twins' answers were less diverse than humans' answers on 154 of them.

The model tended to classify people according to general characteristics, such as age, education, income, and opinions, and to deal with them based on the stereotype of the group to which they belong, rather than simulating the individual characteristics of each person.

The digital copies also showed greater confidence in technology compared to real people, and were more receptive to the use of artificial intelligence systems in evaluating job applicants. They also considered that using people’s data online to build profiles about them was less invasive of privacy.

"distorted mirror"

Ironically, one reason the simulation failed was that the AI was more knowledgeable and rational in some tests than the humans it was supposed to imitate. Humans make mistakes, fall into biases, and may make irrational decisions, while the model tends toward correct and logical answers.

The results indicate that building a faithful digital copy of a person requires simulating what he knows and does not know, even his mistakes and contradictions.

Despite hopes to use these twins in the future in social and psychological research and testing consumer behavior, the study concluded that, in their current state, they are closer to a “distorted mirror” of the human being than to a true copy of him.

Engineers at the Georgia Institute of Technology (Georgia Tech) have developed a new system for communication between wearable medical devices and implants implanted inside the body, by using the body's own tissues as a channel to transmit signals. The system, which the researchers called “Swans,” allows a sensor in one part of the body to send a signal to a therapeutic device in another place, without the need for direct contact or proximity between them.

The study, published in the journal Science, focuses on building a network of small devices that can work together inside the body and on its surface, such that it detects a biological change in one location and then triggers a therapeutic response in a different location, such as stimulating a nerve or operating an implanted device.

The basic idea is to take advantage of the natural ionic conductivity of the body's tissues, instead of relying on traditional wireless technologies, such as Bluetooth or NFC.

Researchers say that traditional wireless signals do not transmit very efficiently through the body, and they require additional antennas and power, which increases the size of the implants and limits the duration during which they can operate.

SWANS sends small electrical pulses through the tissues, and each implant can be programmed to respond only to pulses with a specific voltage and time, allowing the signal to be directed to the desired device within the network.

The team believes that this structure allows the sensing site to be separated from the treatment site. Instead of placing the sensor and therapeutic device in a single unit, each can be placed in the most functionally appropriate location, and signals can then be passed between them through the body's tissues.

Implants less than 3 millimeters in size

Reduced power requirements have greatly reduced the size of the devices. According to the researchers, implants that are less than 3 millimeters in size can be manufactured, making them small enough to be implanted using a syringe.

These devices rely on passive electronic components that consume a very small amount of energy while waiting, and then wake up as soon as the required signal arrives to carry out the specific task.

In tests, the team estimated that a small therapeutic device, operated once a day, could operate for about a year before needing to be replaced. The researchers also reported that the electrical pulses used did not cause damage to the tissue samples they tested.

Motion control experience

The team tested the system in a rat experiment, connecting a network of sensors and neural interfaces across the body. A sensor was able to detect the movement of one of the front legs, then sent a signal through the tissue to another device, which stimulated a muscle in the back leg and caused it to contract. Trying to mimic part of your natural walking pattern. The experiment also showed that a number of devices can work within a single network, without the need for direct wires between them or precise alignment.

Small data inside the body

“Swans” is not intended to transmit large amounts of data across tissues. The design is intended for small, simple signals, such as determining the presence or absence of a specific event or sending an on or off command.

Larger analysis operations, or the exchange of large amounts of data, can be carried out by an external wearable device that acts as a hub for the network, collecting implant readings and deciding when to activate a specific treatment procedure.

This means that the system does not attempt to replace all forms of wireless communication, but rather dedicates a low-power channel for coordination between implanted and wearable devices.

Towards coordinated electronic therapy

The team sees a potential future use in building automated therapeutic systems distributed across the body, where sensors detect biological changes and then trigger an appropriate response at the desired time and place.

The goal is to allow sensors to be placed where they are best able to pick up a biological signal, and stimulators or therapeutic devices to be placed where they are most effective, without having to be adjacent, said Alex Abramson, an assistant professor in the College of Chemical and Biomolecular Engineering at Georgia Tech and lead author of the study.

The researchers indicate that this type of network may support applications in personal electronic medicine in the future, where several small devices work in a coordinated manner, instead of relying on a single implant that performs all functions.

A network that mimics the logic of the nervous system

The team was inspired in part by the way the nervous system coordinates the work of different parts of the body, with a distinct difference in how it works. The researchers point out that the ability to connect multiple sensors and stimulators distributed in different places, including sites deep inside the body such as the stomach, may open the way for more specialized therapeutic systems that are less dependent on large devices or external connections.

The technology is still in a research stage, and has been proven in laboratory and animal experiments, and is not an approved therapeutic system for human use. The next steps focus on developing the network and testing how a larger number of devices work together, leading to applications where sensing and treatment can be coordinated more independently within the body.