Artificial intelligence (IA) can be defined as a system that learns from data how to carry out operations, even in the absence of defined rules. There are two discriminating factors: the need for data and the optional existence of rules. This is enough to declare that not everything that is defined as an AI system is really. Robotic Process Automation (RPA), on the other hand, performs tasks by responding to predefined rules. Tasks that he does very well, but he is not AI.
To understand why the two terms are often used as synonyms and why, even in Italy, there is an abuse of the very concept of AI, we asked for the support of Alessandro Piva, Director of the Artificial Intelligence Observatory of the Politecnico di Milano.
Analyses
Conscious deep learning is still a utopia
by Andrea Daniele Signorelli
The differences between AI and RPA
“Artificial intelligence is closely related to the concept of machine learning, that is the ability of the AI system to improve its performance based on the data with which it is trained, using typical human skills. On the other hand, the RPA is connected to the theme of the automation of predictable tasks, to the implementation of certain actions by the system, on the basis of predetermined and preconfigured actions. For this reason, RPA is well suited to use cases characterized by tasks of limited complexity and of a purely operational nature ”, explains Piva.
One of the many advantages of AI is their ability to deal with problems without analytical solutions. In support of this, among the peculiarities of artificial intelligences is the ability to act where the interaction between a system and the user is not optimized. “AI, explains Piva, is based on learning and therefore data, in reality, are the basis on which artificial intelligence builds its ability to provide adequate answers. The availability of a large amount of data is an enabling element for learning the artificial intelligence system, the important thing is that these data are of high quality and do not contain bias. A discriminating element between the use of AI and RPA lies in the complexity of the task to be performed and in the degree of unpredictability of which the task is characterized. In fact, in these cases, an RPA solution becomes impracticable, due to the number of possible scenarios and the depth of the interaction “.
The analysis
Is it possible to fall in love with an artificial intelligence?
by Francesco Marino
A few examples
Google’s image recognition is AI: it recognizes a zebra because it has been trained by having it examine billions of images and there is no predefined rule that teaches it in which photograph is a zebra or not. Some of the chatbot with whom we talk online, typically to interact with a help desk, are RPAs. There is a group of people who work to identify questions asked by users for the bot to return appropriate answers. But, Piva points out “For example, artificial intelligence, even if we do not realize it, is applied in many contexts of our online experience: the recommendation systems that help us choose services and products (from media, to entertainment, from tourism to financial services) are often characterized by algorithms, which learn our tastes and interests, profile us and offer us further opportunities. Even the aforementioned chatbots, if they are able to learn from interaction with users and autonomously improve their skills, interacting in natural language, can fall into the field of AI. An example of RPA can be linked to the application in the administrative management of orders and payments, where typically there are standardized tasks, very repetitive and with a limited number of exceptions “.
The affinities
We call AI what is not always AI. However, it must be said that both AI and RPA can relieve the human being from repetitive, limiting and even risky tasks. “It depends on the contexts, certainly RPA solutions are characterized by an ‘easy’ business case to present to the company’s top management: they promise to obtain tangible savings from the automation of low value added tasks, repetitive activities carried out in back office contexts. In this case it is possible to free up time for collaborators who can be employed in activities of greater value. Artificial intelligence can potentially be used in an infinite number of business processes and contexts of use and today sees a widespread interest, for the innovation opportunities that can be achieved, not only in the optimization of business processes, but also in the creation of new business opportunities and markets “. In addition, IA and RPA can carry out their respective tasks non-stop, literally up to 24 hours a day, 365 days a year.
Automation
Dawn of the cobots, robots that work with man instead of in his place
by Vittorio Emanuele Orlando
Abuses on the decline
There are companies that claim to use AI in their processes or to equip their products or services with it. These statements are not always truthful but, however, demonstrate the ferment and expectations towards artificial intelligence, as Piva specifies: “We are in a moment of particular media hype, in which talking about artificial intelligence is fashionable and attracts consumers and potential customers. This particular phase of attention, in the technological field, typically precedes a phase of affirmation on a large scale, with concrete and valuable applications. From our Observatory perspective, and therefore from a privileged point of view, we already see the concreteness of the market today, the most advanced large organizations are developing significant projects and even the market for AI solutions startups sees realities of particular interest “. The number of companies that abuse the term AI is therefore decreasing, this coincides with a more precise knowledge of AI, even if only theoretical, by entrepreneurs and financiers.