Six Things We Might Need For Pervasive Computing

There is little doubt that digital technology will become more pervasive than it is even now in the coming decades. Organizations like the Exponential Group argue that digital should be the first step in sustainability, estimating that hardware and software could help reduce emissions by 15% by 2030 and beyond by helping fine-tune buildings, factories, and other environments.

Cars—already packed with processors—will become datacenters on wheels with the growth of EVs, better ADAS systems, and autonomous driving. Health care and remote medicine through new wearables or medical devices is often cited as the largest opportunity for electronics technologies.

Nonetheless, creating an effective and economical ecosystem for pervasive computing systems will require extensive trial and error. As a futurist in Arm’s R&D organization, it’s my job to look ahead and map out the potholes. Taking some of the futurist scenarios as guidelines, here are some of the intriguing barriers I believe we will have to overcome. (And, of course, I would also appreciate your feedback and suggestions in the comments section.)

1. Intelligent tattoos. Neuralink, among other companies, has launched an ambitious effort to use brain and nerve impulses to allow humans to connect to computers. The vision is compelling—imagine how the world would change for people afflicted with limb loss or debilitating diseases—but fraught with grave concerns.

Implanting processors directly into the brain or onto synapses creates profound medical and patient risks. Imagine the complexity of some upgrades. At the other end of the spectrum, computer vision systems that analyze eye motion or voice timbre will inherently be limited: they can only take advantage of a limited set of external data.

Intelligent tattoos would be capable of relaying data to the cloud for analysis or performing AI directly. They could also serve as a gateway for data coming into the brain.

Data integrity and system security will be safety critical. Techniques for preventing DDoS attacks will be needed along with an automated pause button to inspect incoming and outgoing data or guard against irrational impulses. AI algorithms and a human/machine interface for determining true intent would also have to be devised: one could imagine three quick blinks or another simple bodily movement that would not ordinarily be confused for a twitch becoming the new mouse click.

Additionally, it should be easily removeable. Arm recently released a prototype flexible processor and a printed and flexible neural network. While it remains in the experimental stage, the ecosystem for technology components, manufacturing toolsets and software will likely start to gel in the coming years.

2. A digital chain of custody. As UC Berkeley’s Hany Farid has shown, deep fakes of videos, images, or even someone’s voice, are getting more difficult to detect, more pervasive and more insidious. Elections could be won or lost in the future with a few select fakes to create doubt.

Now imagine the potential for mischief in the metaverse. AI-enhanced video calls could be transformed into completely fabricated, and completely convincing, conversations and used to modify your normal behavior or decision process. In the physical world, industrial equipment could send messages that cause employees to shut production erroneously or, worse, not be able to act against catastrophic failures.

For autonomous systems (smart plants or cars), in case of accident, each element of the AI based decision making process, at all levels, should be able to tell you what happened and why a decision was made (e.g., the red light sent a wrong signal, LIDAR did not detect something, etc.). Lieutenant General Vincent Stewart terms computational falsification a fifth-generation of warfare that works by removing someone’s ability to make rational decisions.

In both the physical and in the digital world, we need to be able to trace information to a single point of truth in an immutable way. A data attestation service and blockchain-like ability that scales easily to trace back information to the original bit will be necessary. While such a system may not be able to prove something is true, it could identify intermediate tampering and ensure data integrity.

3. Dissolving ICs. Humanity celebrated another dubious milestone in 2020: for the first time, the mass of human-made materials likely surpassed the amount of natural biomass, and it is doubling every 20 years. Cradle-to-cradle manufacturing, where manufacturers can reuse old materials or parts, can help reduce landfills.

But how do you develop, deploy, and recycle intelligent sensors/systems in a sustainable yet scalable way? Dissolving ICs would give producers a workable way to recover components. Programmable dissolvables could even let them adapt product functionality and aesthetics, thus enabling mass customisation in a scalable yet sustainable way.

4. Data centers in the sky. Although the power consumption of datacenters and networks has remarkably stayed relatively flat over the last ten years, innovations will be required to keep the track record going. Digital data continues to double every two years and AI and 5G will amplify workloads.

Luckily, many pervasive applications won’t require hyper latency. Cold (and lukewarm) data storage and moderate computing loads could conceivably be moved to nano satellites. Although complex calculations regarding total energy consumption would be required, orbital datacenters would have a systemic advantage over their earth-bound counterpart: cooling would be free. It’s a mega-engineering problem to be sure, but it’s also one where a lot of the basic knowledge already exists.

5. AI generated HW. As all industries are moving through their digitalization, electronic industries will have to follow. The design of IoT systems has already been benefiting from application enablement platforms, allowing quicker and easier design of IoT devices and associated applications. About 10-20 of these platforms reach a reasonable level of maturity in seven years. Similar enablement technologies and tools could be developed for almost every segment of the electronic industry where hardware needs to be co-designed with software and applications.

For semiconductors, simplifying and automating the design of complex intelligent systems through some sort of semiconductor design enablement platform could come earlier than we think and simplify developer life. Once massive databases of design data are available in the cloud, layering an AI capable of generating hardware and software, better and faster than human, is a step away.

Neural Architecture Search ML tools could automate the creation of hardware-aware trained neural networks optimized for a given ML task. Similarly, if we can represent hardware components and their relationship in the form of cost functions for a given process, we should then be able to automatically optimize and synthesize hardware for tasks.

6. Phase change memory for things. In 1970, Gordon Moore predicted in Electronics Magazine that phase change memories could hit the market within a decade. It didn’t happen. Traditional memory and storage proved to be more flexible and enhance-able than even their staunchest advocates believed and a then-newer concept—flash—fit well within the traditional semiconductor cosmology. Phase change devices, meanwhile, proved difficult to bring from prototype to production. (Moore shouldn’t feel too bad: another author wrote a piece titled “The Big Gamble in Home Video Recorders” in the same issue.)

A pervasive internet of things, however, changes the equation. Desks, windows, doors, but also a number of durable or non-durable goods will soon start to be augmented with new functionalities, thanks to hybrid electronic systems, and become pervasive HMI or sensors. But they will not be persistently connected or plugged in. They also won’t contain traditional computer interfaces nor battery. Instead, they will be covered in an intelligent second skin that responds to RF waves, heat, or other stimuli. There is definitively a need for low power, non-volatile storage. Could phase change provide the capacity, non-volatility, and power consumption profile needed?

A futurist’s goal is not to predict the future, but to create a view of possibilities about the future and how they can emerge from the present. I’m sharing here a view. What’s yours?