We are on the eve of a new era: the age of Pervasive Intelligence, US consultants David Schatsky, Aniket Dongre and Jonathan Camhi report on Deloitte Insights. Artificial Intelligence (AI) will increasingly be integrated into machines and equipment which can independently identify images, sounds and other patterns without connecting to the cloud, and respond accordingly. This facilitates new applications in a variety of sectors and industries.
The use of AI applications is already widespread in the industry: self-learning systems and similar technologies are now commonplace, especially in production, logistics and distribution. What all these applications have in common is that they are linked to the cloud – the only technology with the capacity to allow self-managing systems and robots to operate.
In other words: there’s no intelligence without internet connectivity – and no intelligence without the drawbacks of such connectivity, namely the risk of service interruption and delays caused by data being transmitted to the cloud for analysis. But if we are to believe Deloitte, these cons will soon be a thing of the past, as a growing number of the new generation of AI applications can run on a standalone basis – a trend known as ‘Pervasive Intelligence’.
Learning from experiences
A key driver behind this trend is that a new generation of AI chips will become available that combine immense processing power with very low energy consumption, Deloitte states. The level of innovation in this area is impressive: earlier this year, for example, MIT researchers presented an AI chip which, depending on the task at hand, is three to seven times as fast as the current AI chips, while requiring 95% less energy. This is crucial, as it means that these types of AI chips are suitable for all sorts of IoT applications not connected to the power grid (e.g. vehicles, drones, cameras, etc.).
As Deloitte predicts, this means that, within one generation, we will be surrounded by machines and equipment that can independently identify – and respond to – images, sounds and other patterns without actually needing to connect to the cloud. In the future, machines will perfect their ability to learn from experiences, adapt to changing circumstances, and predict results. Some devices will be able to deduce the needs and requirements of human users and will even be able to interact with other devices by exchanging information, dividing duties and coordinating actions.
Swarm intelligence
Deloitte regards this latter development as particularly interesting, as it facilitates the application of ‘swarm intelligence’: systems consisting of interconnected, intelligent devices that can interact with each other in order to generate more speed and efficiency. One example of this type of system: smart warehouse robots that can communicate with each other and process pick orders together. This makes it possible to radically reduce pick times for online grocery orders from hours to minutes.
Another example: smart wind turbines that interact with each other in order to generate a maximum amount of energy. Interconnected wind turbines which are equipped with sensors and use algorithms can share information with each other on wind conditions and can change the speed and angle of the rotor blades in real time. This makes it possible to optimise power generation for each turbine, based on continuously changing weather conditions and the operation of the surrounding turbines.
Strategic implications
As you might expect from a leading consultancy firm, Deloitte also reviewed the strategic implications of Pervasive Intelligence. These turn out to be substantial, as Pervasive Intelligence does have disruptive potential. Various types of manufacturing companies may be faced with competition from newcomers offering smart, AI-based alternatives.
Either way, it will take many years for the Pervasive Intelligence trend to start having a significant impact on most sectors. But the ultimate effects will be dramatic, since devices with built-in intelligence facilitate all-new levels of performance and efficiency (with driverless vehicles showing the most potential). Deloitte therefore believes that companies must start now with identifying the potential impact of Pervasive Intelligence on their company and industry, since this is the only way they can carve out a successful position for themselves in reaping the benefits of this technology.
This is a summary of an article titled ‘Pervasive Intelligence’ on Deloitte Insights.
Conclusion: a challenging new field
Robbrecht van Amerongen, Head of IoT at Conclusion, says he recognises the trends outlined by Deloitte: “It is correct that hardware providers are supplying increasingly powerful sensors and standalone hardware modules, also known as ‘edge technology.’ They enable very powerful analyses at the local level, and make decisions there without the need for a call-back to the cloud. But we’re also seeing a growing demand among our clients for fast local decisions to detect irregularities or prevent errors. Applications involving analyses of images, sound, vibrations and power consumption, in particular, require considerable processing power and, since recently, can be implemented at the local level.”
Van Amerongen maintains that the current method we use to manage our data is based on an outdated principle of centralisation. “That may have worked perfectly well for decades, but we’re now getting to a point where we can no longer process all data through a single channel. The amount of data is mind-boggling: the equivalent of 90% of all the data ever produced in the world – that is, since the invention of writing – has been generated in the past two years. And since that rate is only going to increase, decentralised data processing is the future: the central cloud system is bound to get backed up at some point as a result of large amounts of updates.”
Security and privacy
As Van Amerongen explains, there are numerous benefits to local data processing and the use of AI. For starters, there is the speed of decision-making, particularly if these decisions need to be made in real time – that is, within several milliseconds. “If you use a cloud application for this purpose, you’re likely to get a delay of a few seconds,” Van Amerongen says. An added benefit is that the availability and reliability of local AI decision-making are significantly higher, because local AI also remains operational when the system is down or the network is slower than usual. And then there are various security and privacy issues. Van Amerongen: “Some data is not even permitted to leave the production facility or the country. By using local intelligence, you work based on security and privacy by design. The data will not leave the local computer under any circumstances, and only the results are shared with the cloud.”
Still, local AI does have its downsides as well. For one, the intellectual copyright on algorithms and AI modules is more difficult to protect if they run on a piece of hardware at a client’s office or consumer’s home, as opposed to being stored on a well-protected central cloud system. For this reason, local AI is also more vulnerable to hacks, causing the application to behave differently (that is, other than intended). Another issue cited by Van Amerongen is liability: “How are we going to deal with possible losses resulting from decisions made by local algorithms?”, he asks. “This is a huge problem, especially for systems that need to make split-second decisions and where there is a risk of physical injury – as with driverless vehicles, for example. Legislation and case law are light years behind in these cases, since this is relatively new territory.”
Increased expertise
Yet Van Amerongen still sees plenty of new opportunities for Conclusion, an NPM portfolio company, when it comes to local AI. “We are already supplying intelligent edge sensors and modules to some of our clients on a small scale. Demand for a serious, professional and reliable provider that will manage, monitor and update these sensors is only set to increase. For us, this represents an extension of our expertise in data analytics, machine learning and AI because the number of applications will only increase. I consider this a particularly exciting new field.”
Also read: ‘IT service provider Conclusion: How to make big business of big data’