Artificial Intelligence (AI) is a discipline in the IT area that deals with studying the theoretical foundations, methodologies and techniques that allow the design and creation of “intelligent” hardware and software systems. In this period, the success of Deep Neural Networks and large generative models (such as ChatGPT) risks making us lose sight of the true scope of this discipline. In the perception of non-experts we are finally faced with Artificial General Intelligence (breaking latest news), i.e. a system capable of tackling and solving any problem in any situation. Let’s say right away that things are not like this: breaking latest news is still very far away. However, it must be admitted that the results we are achieving in the implementation of AI systems which, in narrow and defined areas, can equal and often exceed human capabilities are truly remarkable.
This means that beyond the enthusiasm of the media (and the market), which focuses on the latest innovation in the sector, AI is based on a very vast set of different paradigms which, appropriately integrated and employed, allow us to address many situations through fruitful human-machine collaboration.
In other words, just as the microscope and the telescope in the past allowed us to overcome the limits of our senses and discover new worlds in the field of the infinitely small and the infinitely large, today AI applications promise (and in part already allow) to extend our cognitive abilities and to analyze and grasp phenomena that otherwise escape our senses and our mind.
What AI can do today
AI applications, with their different paradigms, are today capable of tackling different types of problematic operations.
Feel. There are applications capable of identifying a sound, understanding spoken language and recognizing objects and situations in a video. Since 2017, through controlled challenges, many of these systems have been seen to surpass human capabilities.
Learn. AI applications are able to extract regularities from examples. In this way they learn to recognize a cat from a dog, but also to classify a tumor compared to another form of anomaly by examining x-rays, ultrasounds, etc.
Reasoning. Using methods that we hear little about today, but which instead constitute an important basis for their success, AI applications express reasoning capabilities superior to ours with regards to problem solving, the optimization of processes and scarce resources, the decision support.
Express creativity. On the creativity front, there are AI applications capable of creating innovative solutions, patents and artistic works in musical, photographic, audiovisual and textual forms (from journalistic articles to poetic compositions).
Abstract. This is the most problematic front, in which the capacity of AI is still limited and research is concentrating huge resources: to face this challenge, in fact, AI must work on an enormous amount of data which requires a quantity of energy that it is not always sustainable to produce.
What can be done in medicine
Let’s start by saying that according to the AI Index Report 2023, in the year 2022 it is precisely the medical sector that has recorded the greatest private investments in relation to AI in the USA. In fact, the use of AI today allows us to support the healthcare sector in many ways: both through tools to aid diagnosis and through systems that reduce doctors’ workload in relation to repetitive and bureaucratic tasks. But a more general contribution should not be forgotten, through solutions that make it possible to improve the health of citizens: and in this regard it should be remembered that Sustainable Development Objective 3 of the 2030 Agenda of the United Nations Organization foresees precisely the challenge of “Ensuring health and well-being for all at all ages”.
The following is a general overview of the contributions of AI in the medical and healthcare fields.
Diagnostic support. As already mentioned, today we have tools that can help doctors make diagnoses, identifying visual patterns with a much higher level of precision than the doctor could do alone. In some cases, this type of tool can allow an initial independent diagnosis, which can then be followed by a necessary consultation with the doctor.
Decision support. AI applications that work with artificial vision, sound recognition, movement analysis, etc., associated with logical reasoning and research, provide support for the doctor not only in diagnosis, but also in identifying therapies .
Creation of new medicines. AI can also help develop new drugs: the ability to simulate interactions of substances with each other and with the organism allows us to speed up the phases of designing new therapies and new medicines (there are, for example, applications that have allowed to determine, with a precision never achieved before, the three-dimensional shape of a protein starting from its sequence of amino acids).
Identification of drug mixes for a specific patient. AI can support medicine towards patient-specific diagnostics and administration. There are applications capable of combining data obtained from analysis platforms of tumor proteins and their modifications to identify the enzymes that produce distinctive signs in malignant cells.
Integration with the Internet of Things. The creation of cheap diagnostic tools, which can be integrated into smartphones, smartwatches and other dedicated tools, make it possible to collect useful data to support not only the doctor but also the patient: through monitoring the quality of the air, the level of environmental noise, heart rate, oximetry and blood pressure can, for example, constitute disease prevention tools.
Process optimization and use of resources. During the COVID-19 pandemic, AI applications were tested that allowed hospitals to optimize doctor and nurse shifts as well as manage scarce resources such as masks, gowns and ventilators. This type of applications can also be used on a large scale to optimize processes and reduce stress for doctors and patients.
Reduction of workloads due to bureaucratic aspects. AI tools can also constitute a valid support for interfacing with information systems, for scheduling appointments, for drawing up questionnaires and much more. Think of the use of virtual assistants that provide automatic responses in call centers, but also of more complex applications that, through semantic tools, allow systems to communicate that would otherwise be non-interoperable (for example, it is possible to define intelligent documents capable of recovering information already provided to other systems, while respecting the patient’s privacy and allowing the integration of data within the Electronic Health Record).
The text is taken from “2030. The sustainability of health. New balances between data, welfare and the NHS”, created by the Roche Foundation in collaboration with Edra.
*Gianluigi Greco is President of the Italian Association for Artificial Intelligence (AIxIA);
*Piero Poccianti is Coordinator of the industrial board of the Italian Association for Artificial Intelligence (AIxIA)