A forward-looking blog post on current trends is happy to take a look back at the past year. We would like to analyze and evaluate expectations, future forecasts or new movements that have emerged from nowhere in order to be able to venture an outlook for the coming year.

Looking back at the trends of 2020 , there have been a little different priorities for the future, from which current trends can be derived.

The World On The Edge

This means no forecast of the current state of the world. Rather, it is a reflection of the growing dynamics in the direction of computing “on the edge”, ie the decentralized data processing of a network. Billions of devices are already connected to networks and this number is continuously increasing – which is nothing new in itself. However, the type and requirements of these devices have changed, which has some effects.

Simply put, more “things” connected to a network require more capacity for immediate processing. Autonomous vehicles are an example of this: Whether in relation to communication with the outside environment (e.g. traffic lights) or by sensors that recognize risks (e.g. an object that runs in front of the car), decisions have to be processed in a split second. The latency of data sent from the vehicle over the data processing and analysis network before being returned with a decision on what to do is unacceptably long.

The situation is similar with video surveillance: the current goal should be to promote proactive action at the moment of an event in order to avoid incidents, and not reactively if an incident has already happened. Data processing and analysis must therefore take place within the camera itself.

But the increase in security-related devices in the area of edge computing has a number of consequences.

Computing power in dedicated devices

Dedicated and optimized hardware and software – developed for the specific application – are essential for the transition to a higher level of edge computing. Connected devices require a higher computing power and have been designed for chip use right from the start, which is why Axis is constantly investing in the further development of its own chip. This enables us to design an integrated unit – or a “system-on-chip” – specifically for the video surveillance needs of today and tomorrow. The new generation, the ARTPEC-7 , is also designed with a security-first mentality.

The concept of embedded AI in the form of self-learning systems will also come to the fore. AI – or rather machine and deep learning – has meanwhile gone beyond the catchphrase and has become an everyday reality. It is therefore no longer a trend topic, but has become an integral part of new technologies. In fact, AI will be used even more in the coming year and will find wider acceptance in a wide range of areas and industries than most people suspect. Here too, however, it will be important to develop new models for deep learning in the future that have less data mass and therefore require less storage space and computing power.

Towards the “Trusted Edge”

Trust takes many forms. In the future, consumers will have to trust even more that companies handle our data responsibly. They also have to trust that devices and data are protected against cybercriminals. And also that the data passed on is correct and that technologies work as planned.

Trust in the entire supply chain will also be crucial. While embedding spy chips on the hardware itself is a fairly distant scenario, it is far easier to infect a device with a later firmware upgrade.

Privacy issues will continue to be discussed worldwide. While technologies such as dynamic anonymization and masking to protect privacy can already be used using edge computing, attitudes and regulations are inconsistent in individual countries. The need to go through the international legal framework will continue to exist for companies in the surveillance sector.

With the increasing processing and analysis of data, cybersecurity is becoming more and more important. Even if the increasingly numerous and sophisticated cyber attacks are known, many companies still fail to perform the most basic firmware upgrades. The management of individual devices as well as comprehensive lifecycle management of the entire monitoring solution are of fundamental importance for a secure system. And this is only possible through clear hardware, software and user guidelines.

Regulation: Use Cases vs. technology

It is difficult, if not impossible, to regulate technology (and rightly so, in most cases). It is only realistic to regulate the use cases. For example, facial recognition, which in some applications – for example at the airport so that travelers can be registered more quickly – is considered harmless and even desirable. However, when it comes to monitoring citizens and social credit systems, it is considered to be much more malicious and undesirable. The technology behind it is exactly the same, but the application is very different.

Attitudes to different applications and regulations vary worldwide. The EU General Data Protection Regulation (GDPR) is one of the best known examples. To protect the rights of EU citizens when collecting, storing, processing and using their personal data, it is one of the strictest data regulations in the world. Other countries are far less strict, and many companies with online services from other regions of the world block their access for EU citizens because they are not GDPR compliant.

Regulations have the difficulty that they cannot keep up with the rapid technological progress. Governments will still adhere to the regulations and will continue to control use cases for the benefit of citizens or themselves. It would be desirable to have a more dynamic approach that would keep the industry moving while still closely examining business ethics.

Network diversity

As a direct consequence of the regulatory complexity of privacy and cybersecurity, one can see a departure from the fully open Internet. While the internet and public cloud services remain part of data transmission, analysis and storage, hybrid and private clouds are increasingly used. An increase in “intelligent islands” can be observed here, on which systems for certain applications have only limited and direct connections to other interdependent systems.

While some reject every step out of openness, the arguments in terms of security and data protection are convincing. In addition, one of the advantages of open data exchange was previously seen as an advance in AI and machine learningconsidered. However, the assumption was that machine learning relies on huge data sets to let computers learn. However, advances mean that network models that have been pre-trained today can be tailored for specific applications with relatively little data volume. For example, Axis Communications is involved in a current project to reduce the false alarm rate: a traffic monitoring model was trained with just 1,000 photo examples and was therefore able to reduce the false alarms in accident detection by 95%.

When looking at the technological trends of tomorrow, it is important to keep an eye on the opportunities and risks that we could all face in the future.