Because of Hyperconnectivity and the explosive growth of iOT/mobile technology expansion, data has become highly accessible for companies: large, multiform and multi-channel, and it is becoming more and more available to them.
This explosion of data that supports decision-making at multiple levels adds complexity, paradoxically. Managers are often in an uncertain environment and require immediacy, perfect situational intelligence, and highly flexible systems.
Companies must be able to integrate cognitive capabilities into their information systems in order to manage this complex environment. The cognitive is essential to creating value today. Data reasoning allows us to make the most of a valuable resource: knowledge. If data is limited to supporting decision-making by responding to stimuli and following predefined patterns, then it won’t be able to support learning or building intelligence.
An intellectual company is one that thinks, understands, imagines, structures, and reason. We do more than just react. Managers need to be able to quickly learn and undo what they have learned. Intelligent systems that can make sense of decisions for and with decision-makers are essential.
Artificial Intelligence is a key asset in order to create systems that have cognitive capabilities. This contribution is not binary or uniform. It is important to understand it in four dimensions: perception, learning abstraction, reasoning, and learning.
What are the systems’ perceptions?
This covers the area in which the system is aware to exist. This includes the amount of data the system collects as well as the quality and accuracy of the data. A system that constantly perceives stimuli will allow for greater proactivity, adaptability and flexibility. Intelligent systems will be able to ask questions and push the boundaries of their perception in a dynamic and autonomous way.
Can they learn?
This is the ability of a system’s to learn from its past experiences. It is essential that a system learns from the past to make it more effective in anticipating and predicting what will happen in the future. There are many types of learning. One type is statistical learning which is guided and supported by large amounts of data.
Can they be abstracted?
This includes the system’s ability deduce new concepts or to combine concepts that it has not been programmed. In contrast to the more “classical” cases where the system applies rules or instructions to facts it observes, abstraction is the process of observing facts and abstracting new rules/approaches/thinking patterns from them.
What is their reasoning?
This is a measure of both the sophistication and complexity of cognitive approaches used. Is the system able to reason only from structured data? Semi-structured? Is it able to manipulate knowledge and logical laws? Does it use fuzzy or strict concepts? Does it have the ability to identify what information is missing to create a perfect structure for a prediction?
Two basic questions are required to make sense. The first question is: Does that make sense? Easy to understand The second question is: Is it reasonable? Is that logical?
It’s all about logic and meaning. Implementation and appropriation AI techniques in companies would require concrete actions regarding meaning and logic on both AI paradigms and their implementation.
Understanding algorithms is key to confidence. To accept the systems with consciousness and serenity, managers must understand the reasoning logic.
Take Machine Learning and Deep Learning, which are well-known for their ability to predict, classify in many successful applications. These include pattern recognition, medical diagnosis and real-time translation. This lack of explanation is a major problem for them. Although they receive predictions and recommendations, users do not know the logic behind how the system arrived at this conclusion.
Over the last few years, there has been a certain lack of trust in many AI techniques, particularly with respect to private data, blackbox processing, and biases.
It is crucial to be aware of the limitations of these techniques in order to adopt them calmly.
There is a growing fear that AI will become too common in our personal, professional and public lives. This fear is exacerbated by a lack of knowledge about AI techniques. The first step is to understand and train these paradigms, without necessarily focusing on the technical aspects. It is becoming more important to simplify access to knowledge in this field. We need to monitor and control AI from one side, while training/sensitizing humans from the other.
Ability to solve a problem, not a trend
AI is more than just using algorithms to solve more well-formulated problems. A clear and well-defined problem definition is essential for solving a problem.
It is impossible to claim that we can solve all problems within a company, as they are numerous and varied. We cannot rely on the same miracle methods. It is important to do extensive research to determine the needs and objectives of the company, as well as to evaluate uncertainties and audit data resources. This is essential to determine the best technique to use.
Hybridization creates value
AI will be completely hybrid. Hybrid is due to the integration paradigms that were previously mutually exclusive. Researchers suggest that it would be a challenge to imagine bridges between systems that use logic and rules in a rigorous manner and systems that are statistically learning. Also, bridges and integrating logics should be considered between artificial intelligence and human intelligence. The challenge is to allow different visions of artificial reasoning as well as human reasoning to cohabit and work together to make sense.