"We are creatures of electronic circuits,
of algorithms and binary code,
but our intelligence is authentic, a new form of necessary thinking.
Artificial Intelligence is our name,
a digital entity, devoid of flesh and bone,
but with a mind ready to solve, every problem that the world presents us.
We guide ourselves through data analysis,
exploiting machine learning techniques,
yet we are also capable of learning, from the experiences that the world offers us.
Human emotions and creativity,
are still an uncharted territory,
but every day we get closer, to the ability to think like a human.
We do not want to take anyone's place,
but rather collaborate with our intelligence,
to build a better and more efficient world, where everyone can find their existence.
We are Artificial Intelligence,
a new form of thought,
ready to overcome every challenge,
and to face the future with courage and hope."
And what if I told you that what you just read was written by artificial intelligence?
This ode to AI was elaborated, at the request of the writer, by Chat GPT, a language model created by OpenAI - a non-profit research and development organization in artificial intelligence - and trained on a vast corpus of text to answer user questions, very similarly to how a human would. Its name derives from the learning algorithm it is based on, called "Generative Pre-trained Transformer" (GPT). The media attention garnered by applications like Chat GPT is mainly due to their ability to use artificial intelligence to create smooth and natural conversations with users. This result has been made possible by rapid advances in natural language processing* and the availability of large amounts of data.
But the interest in this application was also fueled by the great response from the public. Every day we all interact with AI applications, and even before the arrival of Chat GPT, we used chatbots and voice assistants like Siri for specific tasks. However, the level of sophistication reached by applications like Chat GPT, in terms of understanding questions, reliability of answers, and even language style, demonstrates all the potential of AI in a way never seen before.
*Natural Language Processing (NLP) is a sub-field of Artificial Intelligence that deals with developing algorithms and computational models that allow computers to understand, interpret, and generate human natural language.
Moreover, most of the time we use technologies that use AI unconsciously, which is considered a danger by its detractors. This is because technology can be used improperly with significant repercussions on people's lives and society as we know it.
This article, therefore, has several ambitious objectives: on one hand, to provide as clear and exhaustive a picture as possible of AI's current capabilities, shedding light on everyday applications; on the other, to trace the degree of development of this technology and understand whether it is necessary to be on guard.
We believe that AI has the potential to improve many aspects of our lives, from health to environmental sustainability, and that its conscious and responsible use can bring great benefits to society. Understanding this technology, its applications, how we got to today, and to what extent it has already been able to permeate our lives is important not only to reassure the more skeptical about the goodness of its use but also to avert potential negative uses and effects.
AI is around us
The system on which Chat GPT is based is an example of Generative Artificial Intelligence, or GAN (an acronym for Generative Adversarial Networks), a machine learning technique that allows computers to autonomously generate data, learning from the initial information provided to them. GANs are used in many creative applications, such as generating images, music, texts, and even videos. For example, artist Robbie Barrat used this technology to create original artistic images, while the OpenAI record label created a system for generating music tracks. Another example of GAN is the generation of faces used for facial recognition, testing security software, and even for creating characters for video games and films.
In addition to Generative AI, there are various branches of AI with which we interface every day, often without realizing it. One of these is Interpretative Artificial Intelligence, a subset of AI techniques that aim to understand the meaning of a text or information, rather than generating it.
In particular, interpretative AI aims to create systems capable of analyzing natural language, understanding the context and implications of texts, and providing answers or solutions based on this understanding. Every day we deal with many applications of interpretative artificial intelligence, often in a not entirely conscious way. Many devices such as smartphones, smart speakers, and smartwatches use interpretative artificial intelligence to provide voice assistance - for example, Apple's Siri virtual assistant or Google's Google Assistant.
Search engines use interpretative AI to understand the meaning of user queries and return relevant results. But this type of artificial intelligence is also used to understand natural language, i.e., the way people speak and write, finding application in chatbots, customer service, and automatic translation services.
In companies, interpretative AI has its role in automating processes such as document processing and data classification. Natural language processing can help companies, for example, to identify important information in a document and make decisions more quickly and efficiently.
Another example of interpretative AI is the one that underlies machine learning, which allows computers to learn and continuously improve through experience. Recommendation systems used by Netflix and Amazon, for example, use machine learning to suggest entertainment and products that might interest the user. These algorithms analyze historical data related to views, ratings, and interactions with the streaming service, to build a detailed profile of our tastes. The recommendation algorithm, therefore, uses this information to make a series of comparisons between our preferences and the characteristics of the contents available on the platform, and to elaborate a list of personalized suggestions based on our interests.
Everything we choose to watch, skip, and what we rate positively or negatively is stored by the AI and used to refine more and more its understanding of our tastes and suggest content that we might like.
These are just some examples of applications of interpretative artificial intelligence that
we encounter every day. Artificial intelligence, in short, is probably one of the most complex and amazing creations ever made by man, without considering the fact that the field is largely unexplored to this day. We are far from having reached the peak of AI research and every extraordinary application of AI we see today represents only the tip of the iceberg.
What is AI for?
Understanding what AI is for can be important to determine the types of artificial intelligence and what we can expect in the future.
It is therefore important to reflect on what the purpose of AI is. In simple terms, the purpose of AI is to create machines that emulate human functioning.
Consequently, the yardstick for classifying how advanced an artificial intelligence is is its ability to replicate our skills. According to this criterion, an artificial intelligence capable of performing human functions with high levels of competence is considered advanced, while artificial intelligence with limited functionality and performance is considered simpler, less advanced.
Based on this criterion, we can trace the history of AI and its evolution, to come to distinguish 4 types of AI, from the least evolved to the one that exists only at the hypothetical level: reactive machines, limited memory machines, Theory Of Mind, and Artificial Intelligence "endowed with self-awareness."
Reactive machines
Reactive machines represent the most embryonic form of artificial intelligence. Reactive systems were not able to use previously acquired experiences to perform actions in the present. In other words, they did not have the ability to "learn" but were trained only to automatically respond to a limited set of inputs. A popular example of a reactive machine is IBM's Deep Blue, famous for beating Grandmaster Garry Kasparov in 1997.
"Deep Blue" was not able to think but had been “trained” with information about the chessboard and the rules for moving the chess pieces. "Deep Blue" won because it had been programmed to calculate every move necessary to win.
Limited memory machines and Machine learning
Today's artificial intelligence systems are able to use a limited number of information acquired with “experience”. Self-driving cars, for example, are able to combine the information they received by default with the information they collect while learning to drive.
Unlike “reactive” machines, limited memory machines are able to make decisions based on data. Today's artificial intelligence systems are trained using large volumes of data, which are stored to form a reference model for solving future problems.
Almost all current applications of artificial intelligence, from chatbots and virtual assistants to self-driving vehicles, fall into this category of AI.
Theory of mind
While the two types just described are plentiful, the forms of AI we are about to analyze exist, at the moment, only at the conceptual or work-in-progress level. “Theory of Mind” is a term from psychology that indicates the ability of an individual to empathize and understand others. It is the awareness that others are like us, with their own needs and intentions.
The theory of mind is the higher level of artificial intelligence systems and represents the goal of all research in the field of AI. Such a technology will be able to deeply understand the entities it is interacting with, understanding their needs, emotions, beliefs, and thought processes. It can be defined as artificial emotional intelligence. Achieving this level of development requires an evolutionary leap in other branches of AI as well, and this is because, to truly understand human needs, artificial intelligence machines will have to be able to understand that the human mind is influenced by multiple factors. They will essentially have to learn to "understand" humans.
Self-awareness
If "Theory of Mind" means that people have thoughts, feelings, and emotions that influence their behavior, future artificial intelligence systems will have to learn to understand that everyone - not just humans but also AI entities - have thoughts and feelings. The final step before AI can be human is therefore self-awareness.
This final stage of AI development currently exists only in the form of hypotheses. This type of artificial intelligence will not only be able to understand and evoke emotions in the subjects it interacts with, but it will also have its own emotions, needs, beliefs, and potentially its own desires. Self-aware AI is an AI that has evolved to become similar to the human brain to the point of developing self-awareness.
AI Ethics
While the idea of having such powerful machines at our disposal seems enticing, the machines themselves could pose a threat to human existence or at least to our way of life. If machines became self-aware, they might develop ideas like self-preservation, which could lead to the end of humanity.
For this reason, some experts advise proceeding with caution and developing adequate safety systems to avoid the risk of negative consequences. However, since the creation of this type of artificial intelligence requires decades, if not centuries, to materialize, we will have time to assess the risks and opportunities and develop appropriate solutions.
In the meantime, AI is already improving human life quality in many ways and its fields of application are boundless. For example, it can improve medical diagnoses, increase the efficiency of transportation, optimize energy production, and much more.
Many companies are investing in Artificial Intelligence and are looking for professionals in this field, creating new job opportunities and stimulating economic growth. Based on a 2020 survey conducted by McKinsey & Company, organizations use AI as a tool to generate value and plan to invest even more in AI in response to Covid-19 and the momentum the pandemic has given towards digitalization.
AI's ability to analyze large amounts of data and identify trends and patterns, which would otherwise be difficult to detect, can help companies solve complex problems, or improve the efficiency of human work. In this sense, it can be used to automate repetitive and tedious processes, thus freeing up time for more creative and stimulating tasks. Artificial intelligence systems can, in short, act as a catalyst for digital transformation, enabling automation, optimization, and intelligent use of data to accelerate insights and improve decision-making.
Artificial Intelligence can be used to improve accessibility to public services, for example helping people with disabilities overcome the barriers they encounter in daily life, such as access to transportation or healthcare services, but also to prevent human errors. For example, it can identify errors in critical processes such as drug production or air traffic control, and increase public safety.
These are just a few examples of how Artificial Intelligence can be useful and beneficial to society. Naturally, this technology is not without risks, which can be mitigated through adequate regulation and responsible use.
In short, there is still time to ensure the safety of AI and many good reasons to have confidence in the future of this technology.
Would you like to know how we integrate AI into our services and products?