A clear definition of “artificial intelligence” does not yet exist, considering that, even though it has been in existence for more than 50 years, it is still an evolving discipline. To simplify the concept, we could say that AI, or artificial intelligence, is a computer discipline that studies and designs electronic computers, both hardware and software, capable of reproducing actions that seem typical of human intelligence.
The first program that reproduced logical reasoning, the “Logic Theorist“, was presented in 1956 at a conference in the USA. It was able to prove some mathematical theorems (output) starting from certain information (input). Since then, many others have emerged capable of proving increasingly complex theorems. Among the best known is the “LISP” (List Processor), the functional language created in 1958 by John McCarthy, aimed at studying specific equations. Since the 1980s, artificial intelligence has also been used for commercial purposes and crossed American borders, arriving in Europe. Then, with the birth of a new algorithm that allowed learning for neural networks, a new era began that allowed access to both computer and psychological fields.
Today, artificial intelligence applications are widespread and appear even in everyday life. Think, for example, of the ability of a car to brake automatically or not to leave a lane while taking a curve (ADAS advanced driver-assistance systems). Consider a computer capable of identifying the famous person we have in mind, asking no more than 10 questions (skill akinator integrated on NLP systems, natural language processing). These are already existing technologies of more or less advanced artificial intelligence, which are changing the way we live, work and study.
Applications of artificial intelligence
Artificial Intelligence or AI is the ability of a technological system to solve problems or develop typical human skills such as:
Acquiring and processing data or images
This includes recognizing an image or a text (AIR, advanced image recognition), and identifying a fingerprint, which is technology that almost all new generation smartphones are equipped with, so as to recognize our fingerprints or facial features. Many computers, too, start up through facial recognition.
Determine a choice or solution
This entails thinking through data processing and determining a choice or solution. For example, the computer can play chess on its own.
Learning from situations
This implies learning through the analysis of the same situation (input) which, placed in different contexts, will give rise to different final processing (output). This includes machine learning where the search engine directs the advertising proposals (output), adapting them to the searches on the net of each internet user (input).
Interact with the user
Here a mobile phone can interact with a person, answering our questions. These are technologies based on natural language processing that, set on certain skills, interact with humans, such as Siri, Ok Google or Alexa.
Different types of intelligent machines
Today, intelligent machines are designed with three different ways of learning. The distinction between these is linked both to the algorithms used and to the objective for which the machines themselves were designed and is aimed at ensuring the best response to external stimuli.
The machines are designed according to three different models:
In this model, the machine has data organized and encoded in a database which, upon request, will be made available. The machine, based on the information received, will decide which is the most appropriate answer. This model can be used in various sectors, not least the medical one where, by entering the symptoms declared by the patient and by analyzing the data entered in the database, one can obtain one or more diagnostic hypotheses.
The machine has unorganized information in this scenario. Therefore, it will autonomously generate a coding model of the same, proposing the solution, in its best opinion, in response to the question that will be formulated. It will therefore have greater freedom of action.
This model is certainly the most complex. Here, the machine must autonomously improve its learning capacity and, above all, must be able to recognize, through the tools made available by man, the likes of cameras, presence detectors, GPS, and the different peculiarities of the external environment. This includes, for example, cars with support for automatic parking, where the device must detect the space available and the maneuver to be carried out for the correct parking.
Artificial intelligence integrated with automation
Today, more and more Robotic Process Automation platforms are developing or incorporating artificial intelligence solutions, generating integrated situations: RPA + AI. You can verbally address a machine with NLP technology, or natural language processing, both to perform a simple task such as sending an email, and to carry out a complex process, such as starting the scanning of an invoice, proceeding to its insertion in a company accounting system and obtaining, via email, a report of the current month’s situation. This also includes the use of AIR technology, advanced image recognition.
The innovative solutions aimed at automating business processes, with the use of both artificial intelligence and RPA technology, make it possible to replace humans in repetitive activities with automatic systems that also include self-learning logics. Mixed solutions of RPA and AI are also referred to as CPA solutions, or cognitive process automation.
As already mentioned, this is the trend of all the platforms that were born to develop RPA solutions and are transforming into more complex CPA or intelligent automation solutions.
These solutions achieve efficiency in the various areas of intervention, freeing up human resources that can be assigned to tasks with greater added value. This corporate efficiency process must maintain a focus on the ethical and responsible relocation of human resources to new work areas. In fact, these digital transformation projects must always be accompanied by change management consultancy. This is so as to ensure the correct involvement of people in the process and an understanding of the technology and, at the same time, to facilitate the requalification of employee’s work following the implementation of innovative solutions.
Obviously, the optimization of business processes deriving from cognitive process automation solutions is represented by KPIs, or key performance indicators, which measure various intervention activities. Innovation will therefore also require the reorganization of the company by creating new roles and new types of work and facilitating the identification of new flexible working formulas, such as smart working.
The implementation process must, however, be divided into quick win solutions, that is, into sub-tasks of limited duration, from two to five weeks. This will allow the customer to ascertain the effective validity of the solution and, at the same time, the work team to make any corrections required.
It is clear that, while implementing quick win solutions, the union of the different sub-tasks will allow to provide automation solutions for complex processes.