The recent AI revolution is based on a few important pillars: faster computing speed, distributed computing, abundance of data and lower storage costs.
To date, however, there still are some critical steps that are missing to achieve true continuous intelligence.
Continuous intelligence refers to a design in which real-time analytics is seamlessly integrated within a business operation to support decision automation. Data processing can therefore be used to respond to events in real-time.
In order to achieve true continuous intelligence we are still missing a couple of key elements, specifically explainable AI and better graph analytics. Deep Learning uses multi-layers neural nets that have been extremely popular in recent years, particularly for computer vision. However, they are still often referred as black boxes because, for all their powerful effectiveness, we cannot yet completely understand how they work.
Tay, the Microsoft twitter chat-bot, made the news in March 2016 when it was shut down because it had started to use inflammatory and offensive language in its tweets. It is therefore clear that explainable AI is needed not only for specific industries, such as banking and insurance, that are heavily regulated, but also to ensure we do not have unintended responses to new events.
Another important area where more development is needed is in the area of graph analytics. Graph analytics allows us to explore relationships between events and between people and organisations and therefore to more deeply understand their interdependence. This is essential for continuous intelligence because it allows for better responses to a continuously evolving world.
Explainable AI and better graph analytics will be needed for businesses to achieve the goal of continuous intelligence in a way that is meaningful, flexible and does not need constant monitoring.
The other important area that will need to be better developed is, of course, natural language processing.
Natural language processing is carving a growing space for itself, but it is so far used predominantly for basic customer interaction or basic document searches. Its importance will be fully realised when it is able to perform on a par with some of the most advanced current vision recognition systems and will then be employed not only to interact verbally but as a true search algorithm able to recognise specific documents, correctly classify them and interpret them according to their meaning and scope. Natural language processing will then wrap around continuous intelligence by allowing to seamlessly integrate it and control it, and allow businesses to easily interact with it.
Reaching the goal of deploying continuous intelligence does raise concerns and fears. One of the issues with current artificial systems and the worries they generate lie in their possible use as tools to control people. Image recognition, for example, is now so good that it is now possible, in principle, to track anyone using the many security cameras existing. It should be noted, however, that the same concerns have always existed with every technological advance: cell phones, for example, can already be used to track people even when turned off. In the same way that legislation has been enacted in the past, especially in Europe, to protect people’s privacy rights, these concerns can be addressed by passing privacy laws to protect citizens’ rights.
A second issue is around the impact on new jobs. History, however, has taught us that any previous technological advance has produced better labour laws and, rather than killing jobs, has produced different ones and often contributed to reducing the number of working hours. The initial evidence seems to suggest the same is happening at the onset of the current AI revolution.
While many businesses are still struggling to focus and gain advantage from the current AI revolution, it is becoming increasingly clear that moving towards true continuous intelligence will be the key differentiator between the players of tomorrow’s business world.