The future of computing

Billions of CPU chips pour out of chip fabs annually. The large majority are installed in IoT devices where they are targets for botnet hackers. Increasingly popular machine learning is seen as benign, if not miraculous, but ML works for "black hats" as well as for "white hats". Meanwhile, data analytics influence our votes, not just our purchases. And on the horizon looms quantum computing which threatens many of our notions about security and encryption. All is not well.


In the year 2000, people far outnumbered computers. Today, the reverse is true. The human population of the world is about 7.6 billion. Perhaps 6 billion smartphones have been sold and people also own an estimated 2-3 billion PCs. The number of (typically invisible) IoT devices connected to the Internet in 2018 is estimated to be over 23 billion and rising rapidly. Computer chips are installed in all sorts of devices: our cars, watches, Fitbits, home thermostats, TVs, refrigerators, remote door locks, home routers, air-quality sensors, and even smart pill bottles that ensure we are taking our medications. Many if not most of those devices are connected to the Internet. Collectively, they have been dubbed the Internet of Things (IoT for short). The growth of the IoT is driven largely by the chip industry's need to keep the many chip fabs busy as the smartphone market has matured. It also serves to provide increasingly vast quantities of data about every aspect of our lives. Many data predators silently gather, analyze and exploit such data. The IoT also enables Internet blackmail: for exmple a botnet based upon enslaved IoT devices launched DDOS (Distributed Denial of Service) attacks that shut down the likes of PayPal, Amazon, Twitter, and Netflix for several hours on Sept. 21, 2016. In January, 2018, a similar botnet was used for a DDOS attack against three large financial industry targets.

Various players, both corporate and government, gather as much as possible of the vast volume of data generated by all these end-user devices for analysis by the world's server farms. Together, those farms contain perhaps 75 million servers that run 24 hours a day. They spider the web, run various social media sites, help to guide car drivers, serve up music and YouTubes, and gather and analyze data from various sorts of IoT devices. They analyze such data to determine social facts such as where on planet Earth each of us is at any given moment, who we are with, who we communicate with, what we are doing, what we buy, how we think, and most recently, how we vote. Such data gathering was originally designed to guide ad placements and other marketing tactics. Although annoying, we are used to everpresent marketing. But in 2016, Cambridge Analytica together with the Russian GRU (equivalent to the American CIA and NSA together) began using such data to manipulate elections (e.g., the 2016 British "Brexit" election, the 2016 American presidential election, and the 2017 French presidential election). The world's data analytics servers together use perhaps 10% of the worlds electric power. In 2015, they used 416.2 terawatt hours of electricity which far exceeded UK's total electric power consumption that year. And server farm power usage is expected to tripple in the next decade.

The basic phenomena that drive the future of computing for good or ill include:
  • The next generation 5G wireless connectivity from any digital device to any other -- Ericsson estimates the number of 5G subscriptions will reach one billion by the end of 2023. 5G will support both high speed consumer applications and myriad industrial applications.
  • The continued emergence of what might best be called a "Cyborg" culture based on wearable computing communicating with other "Smart" devices.
  • Ever more abundant, more powerful, and cheaper digital chips. "Smart Homes" and cars contain dozens of chips that communicate wirelessly with an Internet hub that is accessible from a smartphone from anywhere in the world. So a smart home becomes a small multicellular digital entity with all the issues discussed in this website.
  • The evolution and deployment of Cyber Warfare (See history of Cyberwarfare) one example being the Russian attack on the 2016 US Presidential Election.
  • The growing number and variety of results from Machine Learning
  • Cryptocurrencies may well roil the waters of world finance.
  • The maturing of Quantum computing will become another disruptive factor in the next decade
  • The eventual successful development of Artificial General Intelligence (AGI) will have unpredictable impacts on society
  • A new digital culture is arising. Call it the "Cyborg" culture. The many new apps and new devices, especially mobile and wearable devices, apple watch have a subtle effect on our minds and on our society. We find that the virtual world manifested in our mobile devices frequently seems to nag us to interact with it like a pet dog that always wants us to play. All too many of us cannot resist that temptation. This growing symbiosis between human minds and social cyber-interactions has spawned a new discipline, cyberpsychology, that studies the many effects seen in social media communities and other virtual digital worlds created by humans for humans that are manifestly addictive or otherwise maladaptive. Such dependence tends to blunt our skepticism so that we are susceptible to social engineering. Hacking and entry into supposedly secure corporate or government systems is susceptible to social engineering attacks such as spear phishing or whaling that exploit our social habits of trusting our friends and co-workers. For example Russian cyber warriors broke into the Democratic National Committee prior to the 2016 Presidential election with spear phishing tactics.

    The digital ecosystem is at bedrock, based on digital chips. They are made in billion dollar chip fabs all around the world: USA, India, China, Costa Rica, Ireland, Russia, Mexico, Netherlands, Singapore, France, Tiwan, Israel, etc. But both computing and society are being overwhelmed by the ever smaller, faster and more numerous chips being produced in these chip fabs. However, Gartner in Feb, 2018 reported the first ever decline in smartphone sales. Where will new markets be found?

    The fastest and most expensive chips typically go into powerful servers in the world's server farms that analyze every bit of data that our activities generate. Roughly 80% of these high-power chips are from Intel. Most of the rest from AMD. Less expensive processors go into PCs and Smartphones although those markets are beginning to saturate.

    The Internet of Things" (IoT) is believed to offer endless demand if the chips are cheap enough because so many devices can be made "smart" simply by including a cheap System on a Chip (SoC): Smart coffee makers, smart thermostats, a doorbell that notfies you wherever you are through your smartphone, smart lawn watering system that track special local weather sites to judge when the lawn needs water, or a smart voice activated ceiling fan switch or smart dog feeder/toy. At the top of the pyramid are the consumers of chips, which used to be corporations with clear economic criteria for buying. Now people of all sorts buy digital devices, including kids using computer games of various sorts, people using smartphones and social media, and people who are enamored of "smart" things.

    There are three important distinctions between top-of-the-line server chips and bottom-of-the-barrel IoT chips: price/volume, security, and work load. First, is price/volume -- IoT chips are cheap and are built in huge volume, PC and smartphone chips are more expensive and are sold in considerably lower volume, and server farm chips are specialized for high compute load and are sold in still lower volumes than smartphones and PCs. Second is work load -- servers are seldom idle, PCs and smartphones are idle part of each day, while IoT chips are almost always idle. And third is security: server farms have very professional security, most smartphones and PCs have middling security, and IoT chips are somewhat like Harry Potter's Cornish Pixies -- if not properly restrained (by secure passwords) and/or have good intrusion detection and automated shutdown (apoptosis), they become ill-behaved DDOS weapons that can, and have, temporarily shut down major portions of the Internet and blackmailed large financial institutions. Few purchasers of the devices containing IoT chips are even aware of, let alone cautious about, the insecurity of their default passwords, let alone knowlegable about how to change the default passwords to sufficiently secure ones.

    Cyber warfare, -- Cyber-warfare is now an explicit tool of some nations, Russia being the most overt player. The USA and China are more circumspect (Photo shows a defensive US Cyber facility). USAF cyberwarriors In May, 2008, Sandia Labs, with funding from Department of Energy, organized and hosted a two day meeting of a wide variety of experts to explore the future of Cyber Warfare and possible approaches to counter or prevent attacks. See Cyberfest Report). History tells us that the Internet became a target for malicious attacks by amateur hackers nearly as soon as it was publicly available. However, large-scale overt cyber attacks on Western democracies became publicly visible only recently with Russia's meddling in the June 2016 British "Brexit" election. At the same time, the Russian GRU began working intensily at affecting the American 2016 Presidential election. "Russian GRU officers hacked the website of a state election board and stole information about 500,000 voters," according to DOJ spokesmen. "They also hacked into computers of a company that supplied software used to verify voter registration information." The defendants worked for two units of the GRU known as "Guccifer 2.0" and "Fancy Bear" that "engaged in active cyber operations to interfere in the 2016 presidential elections," said US Deputy Atty. General, Rosenstein. Twelve Russians have been indicted by the US Dept. of Justice. Their propaganda was directed in large part by the Data Analytics output of a British company, Cambridge Analytica, owned by American billionaire Oligarch Robert Mercer. They analyzed information on millions of Facebook users and provided the results to the Russian hackers.

    Machine Learning is growing rapidly and it too deliver up to 11.5 petaflops of machine learning acceleration can benefit from custom hardware. According to Bloomberg News "...all the big tech platform companies (and lots of startups) are making chips (for data centers) optimized to run the specific maths used in machine learning as fast and efficiently as possible". An ASIC Machine Learning accelerator such as the one pictured here can deliver up to 11.5 petaflops of machine learning acceleration. Machine Learning can be used to automate IoT bot gathering and herding. Such a system could create bot herds with sizes optimized for the particular target, handle the communication with the target, and collect the ransom via anonymous cryptocurrency, all with no visible illegal behavior on the part of humans. Perhaps this is already happening now?

    Quantum computing -- Many schemes for encrypting financial data or political secrets depend upon factoring large numbers (multi-hundred digits). It has already been shown that factorization problems can be solved very rapidly with a quantum computer that has enough quantum bits (qubits). Intel has demonstrated a 49 qubit chip, IBM has built a 50 qubit quantum computer and Google recently announced a 72 qubit computer. While programming such machines remains a largely unsolved problem, it is believed that a 100 qubit machines would be more powerful than all today's supercomputers combined. But before one panics over the possible loss of all useful encryption, it should be noted that not all encryption must use factorization as the "difficult" computation. Other schemes, such as most current symmetric cryptographic algorithms and hash functions, are considered to be relatively secure against attacks by quantum computers. But a wholesale conversion of all factorization-style encryption would be a large undertaking.

    Crypto currency -- Bitcoin mining tends to be done via standalone ASIC Mining hardware or mining pools in places with very cheap power from windfarms or from dams on the Columbia river. But free is superior to cheap. More than half a million machines have been hijacked into one of several cryptocurrency miner botnets. One botnet has mined nearly 9,000 monero tokens (worth roughly $3.6 million). Those who direct botnet crypto-mining pay nothing for the CPU time or the electric power so they can outcompete "legitimate" cryptocurrency miners due to lower costs. Perhaps botnet miners are in part responsible for the fact that cryptocurrencies as a class of assets have crashed dramatically as of September, 2018.

    Adversarial AI -- AI is not just for the "good guys". It can improve and direct whole server farms in attacks. "What's different about adversarial AI attacks? They can put on the same malicious offenses with great speed and depth. While AI is not a fully accessible tool for cybercriminals just yet, it's weaponization is quickly growing more widespread. These threats can multiply the variations of the attack, vector or payload and increase the volume of the attacks. But outside of speed and scale, the attacks are fundamentally quite similar to current threat tactics."

    Just who will protect us from all these new sorts of risk? "According to a recent survey, 66 percent of information security professionals believe there aren't enough qualified analysts in the field to handle the increasing volume of security threats. In addition, many organizations have limited budgets, restricting security teams from hiring the talent they need to protect their networks. AI-powered tools can automate security processes and perform complex tasks, freeing overworked analysts to focus on more pressing matters. ...Time is a critical resource for security analysts, who must determine whether to escalate an alert or write it off as a false positive in under 20 minutes. Due to the around-the-clock nature of incident response, security teams should invest in machine learning tools that can filter out the noise and present reliable analysis with speed and scale."


    Last revised 9/11/2018