There has been a lot of hype around 5G and AI these past couple of years. Newspapers, websites, tv-pundits, and politicians often talk about the potential impact these technologies will have on the world. This is completely justified as 5G and AI represent the next technological revolution that can drastically alter our lives in many ways. 5G and AI have the potential to digitize and automate tasks in the fields of medicine, education, computer science, robotics, banking, governance, warfare, space, virtual reality, and much more. Humans will be able to do things considered fantasy not too long ago. The potential applications of 5G and AI are virtually endless. Moreover, 5G and AI are complementary: they help each other perform better and reach new potentials.
Yet with all the new possibilities and changes of these new technologies a new set of risks emerge that we have to take into account. This article looks at what these new technologies are, what they could mean for humanity, and what the potential risks are.
What is 5G?
5G is simply the next step in mobile telecommunications. 1G wirelessly connected mobile phones to radio towers, allowing you to have a conversation while you walk. 2G turned analog into digital communication, allowing you to send text messages too. 3G made the jump to having the internet on your mobile phone. 4G improved the speed and volume of internet data transfers to such an extent that you could stream whole movies on your phone. 5G is next in line, and it is approaching fast. The next step in telecommunications will allow you to achieve internet speeds up to 100 times as fast as 4G with virtually no processing delays.
5G will make the ‘Internet of Things’ a reality
With 5G, it will also become possible to connect literally millions of internet-connected devices, appliances and sensors without draining their batteries. Such new networks will make it possible to have entire smart homes and smart cities. Such a globalized network of devices, appliances, and sensors is sometimes called the “Internet of Things”, or IoT. It will allow all these devices to transfer data and communicate without any human input.
This “IoT” will enable businesses to automate all sorts of processes, reduce production costs, lower waste and increase transparency. For example, with the right programming, the IoT could create self-sustaining farming units. Soil sensors would alert drones what amount of water or nutrients the crops require for a maximum yield. Cars could be equipped with tire sensors that direct self-driving cars to a mechanic for a timely tire replacement. The possibilities for setting up self-sufficient ecosystems through a vast interconnected web of mutually assissting devices seem to be limited only by our imagination and willingness to invest. It is no surprise that huge economic gains are expected from the introduction of 5G. The Global management consulting firm McKinsey, for example, expects that the Internet of Things will have an economic impact of up to $11.1 trillion by 2025.
5G will help AI break new ground
In order to expand the capabilities of AI we need to feed it more data. A LOT of data. The recent advancements in AI are the result of two developments. First, years and years of improved computing power. This power indicates the number of calculations that a computer can perform in a second. The other development has been the explosion of personalized data that suddenly became available when people all over the world started uploading their personal data on social media websites. The amount of public data that suddenly became available coincided with the ability of computers to actually process that data at lightning speed.
Further improving AI will require new techniques and even more data for computers to sort through. Artificial intelligence requires large amounts of data to train its underlying algorithms. 5G makes it possible to record and transmit far larger data-sets across platforms than is possible right now. In other words, 5G will help overcome the technological barrier of feeding AI algorithms enough data to help AI become more advanced.
In short, 5G has the potential to make the next big technological revolution a reality. The potential for increased productivity is almost unimaginable. Combined with the advancing field of AI, the applications for these emerging technologies are virtually limitless. But what is Artificial Intelligence exactly? How can machines ‘learn’? And what are the possible risks of AI?
What is the definition of AI?
AI, or Artificial Intelligence, is intelligence demonstrated by machines. This means that machines such as computers perform tasks which require some form of intelligence. Another description of AI is: “A machine is said to have artificial intelligence if it can interpret data, potentially learn from the data, and use that knowledge to adapt and achieve specific goals”. The idea behind it is that human intelligence can be so precisely described that a machine can be made to simulate it. This is why AI is often used in connection with robots: machines that are basically copies of humans with the same capabilities.
There are two very different conceptions of Artificial Intelligence. The first ‘type’ or idea of AI is the one you recognize from famous movies like 2001: A Space Odyssey or the Terminator. These are machines or systems that think, plan, and respond just like humans, while also possessing so-called ‘superintelligence’. This is called Artificial General Intelligence (AGI) and would be able to process information at lightning speed, make incredibly complex calculations in nanoseconds, and never forget anything. You can imagine it as Google with its own mind. Right now, no such technology exists. Scientists do not know if AGI is even realistically possible.
The second version of AI is called ‘Narrow AI’. This is the AI that actually exists and is being further developed as you are reading this. Narrow AI are systems that do distinct tasks incredibly well, such as self-driving cars, voice recognition, or software that can make medical diagnoses based on advanced imaging. Within Narrow AI there’s a distinction between different types of learning.
The different types of Narrow AI learning
Within Narrow AI, there are different types of learning. In the table below you’ll find a brief and simplistic description of these kinds of learning.
|Type of ‘learning’||Description|
|Machine learning||Machine learning involves using examples and experiences in data form to refine how computers make predictions or perform tasks|
|Supervised learning||Supervised learning is showing AI labeled example data, like photographs with descriptions, to “teach” a computer how to interpret and categorize them|
|Unsupervised learning||Unsupervised learning means feeding a computer data without any annotated or labeled guidance|
|Reinforcement learning||Reinforcement learning is software that experiments with different kinds of actions it can perform to figure out how to maximize a virtual reward, not unlike scoring points in a video game|
|Deep learning||Deep learning is possibly the most well-known form and potentially the most groundbreaking type of learning. Deep learning allows machines to ‘learn’ by letting it sort through enormous data sets and then recognize patterns, find correlations, and make inferences based on probabilities. This technique has allowed AI to do amazing things such as: beat the world’s best chess player, correctly diagnose melanomas, engage in complicated conversations with humans, drive cars, beat video games, paint portraits, and even make scientific discoveries.|
For the record, the actual mechanics of AI learning are far more complex than described here.
For computers, image-recognition is harder than for humans (more on that below). This is because computers are good at matching zeros and ones but not at identifying objects. AI will easily recognize two identical images of a cat because these images will have the exact same number of pixels (among other properties). However, this does not mean that the machine recognized the cat as a cat. When the same cat is shown in other images, the machine will not recognize it. In order to do this, complex math problems have to be solved through a neural network.
The goal of deep learning is to reverse-engineer the human brain’s learning capabilities. Neural networks simulate the network of neurons in human brains in order to make decisions in a more human-like way.
The possibilities and limitations of AI
Various studies have been released the past few years claiming AI will be a true economic game-changer. PwC Global, a professional services network, predicts that “AI could contribute up to $15.7 trillion to the global economy in 2030. Put simply, there is an absurd amount of money to be made from 5G and AI.
Some of the sectors that stand to gain the most from AI are healthcare, automotive industries, financial services, retail, technology, communications and entertainment, manufacturing, energy, and transport and logistics.
However, it is also important to see the limitations of AI as it currently exists. AI neural networks up until now only have a few million “neurons”. Which is still very little compared to the 100 billion neurons inside every human brain and its trillions of synapses. On top of that, AI neural networks are “modeled” on human brains; yet human brains are so incredibly complex we are still far away from completely understanding them. In other words, AI neural networks are an incomplete imitation of something so complex we have not figured out yet – if we ever will.
To give a simple example of the limitations of AI as it exists today: “a ‘deep learning’ system running on 16,000 processors taught itself to identify cats – with 75 percent accuracy – after analyzing 10 million images.” A three-year-old child can correctly identify cats after seeing two or three during a walk in the park. This form of AI is called “narrow” because, at the end of the day, the AI is only as good as the data it is being fed. Humans still control the input of data and are challenged to come up with complex networks and equations in order for AI to work. Moreover, these deep-learning algorithms are, unlike humans, not able to consider ideas or concepts that they have never encountered before.
In short, there is a lot of potential for these technologies; but we’re still far from reaching that potential.
Common myths about AI
Because the term AI is being used a lot in movies and other media, people have developed some common misconceptions about AI. In the table below we discuss various well-known myths that exist about AI and what the truth actually is.
|‘Superintelligence’ is just years away||The next stages of narrow AI is most likely decades away. AGI might never come to exist.|
|The creation of an all-powerful AGI is inevitable||It may happen. It may not. AI experts disagree and we simply do not know.|
|Only people who are already scared of new technologies worry about AI||Many top AI researchers as well as other scientists have expressed concerns over AI and the direction it’s going in.|
|AI could turn conscious or evil||The more likely scenario is that AI will misunderstand human goals. Telling AI to “get you to the hospital as quickly as possible” could cause a self-driving car to exceed the speed limit and cause many accidents because its only objective is to get there as fast as possible without considering the context. This is a miscommunication between humans and AI leading to misaligned goals. This is different from AI turning evil.|
|Robots are the main concern when it comes to the dangers of AI||The main concern when it comes to AI is actually ‘misaligned intelligence’ as stated above. This is where the goals of the AI don’t match our goals.|
|Machines cannot have goals||Machines actually can have goals. For example, a heat-seeking missile has a goal, namely: to hit its target. The problem that arises is when those goals are misaligned with the goals of the human determining those goals.|
The concerns over 5G
Given that 5G and AI have such enormous potential to change the world, it is only natural there are also tons of concerns. The worst fears center around AI, but 5G also has many vocal critics who fear that the new telecommunications infrastructure will produce negative health effects.
5G uses a higher frequency of radio waves than 3G or 4G. This higher frequency allows for more devices to have access to the internet at the same time. This is what will enable the Internet of Things. However, some people worry that this higher frequency in combination with far more devices constantly communicating with each other will have a negative effect on our health and that of other animals.
This is a reasonable concern. If there is a large increase in radio waves and electromagnetic radiation across all major cities, it would make sense to at least question if this could not have some side effects. But it is important to remember that, up until now, no negative health effects have been linked to the radiation of telecommunication devices, meaning radio towers or mobile phones. In 2014, the World Health organization released this statement: “no adverse effects have been established as caused by mobile phone use”. This does not mean it is impossible that negative health effects can be caused by 5G. But, so far, no evidence has been found.
The risks of the Internet of Things
The Internet of Things will connect more devices together than has ever been done before. This can produce cooperation and efficiency across a number of industries but it also carries a number of risks.
- The larger the network of devices is, the more devices that are vulnerable if a hacker manages to breach the security of the network
- The amount of devices that will be linked together might be too many for mankind to realistically manage. There will come a point where maintaining oversight will be impossible
- If there is a bug in the network, it could be possible that every device in the network will be negatively affected by it
- There is still no global standard for the compatibility of all the devices that should become part of the Internet of Things. This could cause communication problems across the devices
The risks of developing AI
AI especially has many people concerned. The creation of an AGI could possibly get out of hand and spell disaster for mankind, just like in the Terminator. Thankfully, this is the least likely problem to arise from AI. Nevertheless, there are a number of other serious concerns surrounding the development of AI.
|Elon Musk||The global competition over AI dominance could lead the world’s most technologically advanced nations (the US, China, Russia, Germany etc.) to accidentally cause World War III.|
|Stephen Hawking||AI brings many disruptions that could lead to the greatest disaster in human history. Think of powerful autonomous weapons, new ways to oppress the people, or mass economic disruptions|
|AI developers||Algorithmic bias (such as discrimination towards women or minorities) could accidentally be picked up by AI and lead to further discrimination in terms of job opportunities, entitlements, scholarships, and many other areas|
|The European Union||Without a regulatory framework to protect individual freedoms and privacy, AI could be used as a tool for mass surveillance and a weapon against personal liberties|
5G and AI are not around the corner yet. It will take at least a few more years before we can see the wide-ranging effects of 5G technology deployed on a mass-scale. AI is most likely still decades away from reaching the level of intelligence that many people speculate over. 5G provides a much clearer picture of what we can expect in terms of pros and cons. One thing is certain, we will be hearing even more about 5G and AI the next couple of years.