By now terms linked to the development of the super technology of the future, Artificial Intelligence, have forcefully entered our vocabulary, sparking heated debates between experts, politicians, consumer associations and high tech industries. The citizen is disoriented and would like to understand more about this revolution which could soon upset his real life: For better or worse it depends on your point of view.
By Massimiliano D'Elia
Europe was the first to try to regulate withIA Act the use of intelligent algorithms. A regulation welcomed by politicians and experts but which still leaves doubts and requests for further information from consumer associations and other organizations far from political and social dynamics. business.
The preliminary agreement between the Council and the European Parliament, established at the beginning of December, outlined the architecture of the law, but the negotiation between the European institutions and companies leader in the AI sector is still at sea.
The definitive approval of the AI Act is expected between next March and May. Its effects will, however, only be seen as far back as 2025, when the law will come into force with the aim of promoting technological innovation and protecting the fundamental rights of citizens.
The European regulatory framework on artificial intelligence, for now, has been based on an approach "risk-based” with increasingly strict progressive rules for high-risk systems, whose developers have the obligation to register and the need to make the data used in the training of algorithms. Although some European countries had initially tried to protect advanced artificial intelligence models from rigorous rules and controls, the final agreement provided for some limited concessions, such as for national security and therefore for all those activities related to research and research. innovation of military systems and police forces.
Biometric identification and mass surveillance are restricted by authorization by dedicated national authorities. Legal coverage to use predictive technologies will be guaranteed only to law enforcement and security forces. European law therefore imposes transparency obligations for general purpose artificial intelligence systems before their release on the market and more careful management of sensitive data.
The entry into force of the AI Act will be followed by a period of two years for the implementation of the regulation by member states and a further six months to establish prohibited uses. Furthermore, a AI Pact, i.e. a voluntary compliance system that anticipates European provisions.
Fines for violations of the AI Act will range from a minimum of 7,5 million or 1,5% of turnover to a maximum of 35 million or 17% of turnover. In light of the many interpretative variables, the interests of companies and consumer associations, the EU regulation is far from defined and may undergo substantial changes, thus modifying the collaborative spirit that animated the drafting of the initial document approved at the beginning of December 2023.
Artificial Intelligence and its impact on nature, on blue gold
Water consumption in artificial intelligence (AI) servers is a very important aspect when evaluating the overall environmental impact of an AI system or data center. However, it is important to note that water consumption is often an indirect aspect related to various factors, rather than an area of direct use as in the case of energy. Here are some factors that can contribute to water consumption in AI servers:
Data Center Cooling: Data centers, which host servers for processing AI workloads, require cooling systems to keep temperatures at acceptable levels. These systems can use water to dissipate heat generated by the servers.
Hardware Production: Manufacturing dedicated AI hardware, such as graphics processing units (GPUs) or specialized processors, often requires significant amounts of water for manufacturing and cooling processes.
Energy Production: If the energy used to power AI servers comes from sources that require large amounts of water (such as coal-fired or nuclear power plants), the full lifecycle of the system can impact overall water consumption.
Hardware Manufacturing Resources: Extracting natural resources needed to produce hardware can impact water, especially if it involves materials like silicon.
Geographical Location: The availability and management of water resources can vary greatly based on the geographic location of data centers. In regions with scarce water resources, water consumption can be a critical issue.
Some companies are adopting strategies to reduce the overall environmental impact of their AI systems, including water consumption. This may include adopting more energy-efficient technologies, employing innovative cooling systems that require less water, and transitioning to renewable energy sources. It is important to note that environmental considerations related to AI, including water consumption, are becoming increasingly relevant, and many organizations are looking to implement sustainable practices in the design and management of their AI systems.
THE DECALOGUE
Below is the explanation of some terms that have forcefully entered our vocabulary and which could disrupt our real lives faster than emerging legislative regulations.
Artificial Superintelligence
“Artificial superintelligence” refers to an advanced level of artificial intelligence (AI) that significantly surpasses human cognitive capabilities in several areas. This term is often associated with AI that excels at diverse intellectual tasks, machine learning, and solving complex problems. Artificial superintelligence is a futuristic vision that suggests the creation of intelligent systems that not only surpass human learning and understanding capabilities, but that can also develop an autonomous understanding of the world, reason, learn, and solve problems in ways that are beyond the reach of human intelligence.
Today we have not yet reached a level of artificial superintelligence, today's research and development focuses on artificial intelligence systems that can perform specific activities or tasks more efficiently or precisely than humans. Artificial superintelligence is more of a future prospect that deserves deeper ethical reflection on the possibility of creating entities with an intelligence capable of surpassing that of humans.
Generative Artificial Intelligence
“Generative artificial intelligence” refers to a subfield of artificial intelligence (AI) that deals with the creation of systems that can autonomously generate new data, content, or information. These systems use machine learning approaches, particularly generative neural networks, to produce data that often mimics or is indistinguishable from that generated by humans.
An example of generative artificial intelligence is the concept of “Generative Adversarial Networks” (GAN), in which two neural networks, a generator and a discriminator, are trained simultaneously through a competition process. The generator tries to create increasingly realistic data, while the discriminator tries to distinguish between real data and generated data. This competition leads to the continuous improvement of the generator's ability to create increasingly convincing data.
The applications of generative artificial intelligence are diverse and include the generation of images, texts, music and more. For example, GAN can be used to create artificial human faces that look authentic or to generate landscape images that look realistic. However, it is important to note that the use of such technologies also raises ethical questions, such as the possibility of manipulation of content or the creation of false information by international actors such as terrorists of different origins.
General Artificial Intelligence
Artificial General Intelligence (AIG) represents a level of artificial intelligence that aims to understand, learn and perform any human cognitive activity in a similar or even superior way to humans themselves. Contrary to normal AI, which is designed to solve specific or limited tasks, IAG aspires to a broader and more flexible form of AI, capable of adapting and learning across multiple domains.
The key characteristics of Artificial General Intelligence include
General Learning: The ability to learn from a wide range of data and apply this knowledge in different contexts. This type of learning goes beyond the specific learning of simple tasks.
Analogical Reasoning: The ability to solve complex problems and make connections through analogical reasoning, similar to the way humans approach new situations based on past experiences.
Understanding the Context: The ability to understand the context in which one finds oneself, considering environmental, social and cultural factors. This skill is essential for adapting to new and unexpected situations.
Self-awareness: Awareness of one's existence and capabilities, including recognition of limitations and the ability to learn from mistakes.
Adaptability: The ability to adapt to new tasks or environments without significant reprogramming.
Currently, IAG is one of the most ambitious and complex goals in artificial intelligence. Most current AI technologies are specialized in specific tasks and do not possess the flexibility and cognitive breadth that would characterize a general artificial intelligence. Achieving artificial general intelligence raises significant technical, ethical, and security challenges and remains a long-term goal in AI research.
Sensory neurons in AI
In the field of artificial intelligence, “sensory neurons” can be associated with components of artificial neural networks designed to process data from sensors, similar to the way the human nervous system uses sensory neurons to perceive external stimuli.
In an artificial neural network, the term “neuron” refers to a computing unit that receives inputs, processes them through an activation function, and produces an output. In the first layers of a neural network, which are often called “input layers” or “sensory layers,” neurons are responsible for receiving and transforming initial information from sensors or input data.
For example, in a computer vision application, sensory neurons can represent the raw input from pixels in an image. Each neuron in this initial layer can be associated with a specific pixel and can be trained to respond to certain patterns or features in the image. In other words, these sensory neurons learn to recognize certain aspects of images, such as outlines, colors or shapes.
Sensory neurons are just one part of a larger AI model, and their output is then processed through successive layers of the neural network. The ultimate goal is for the network to learn increasingly complex and meaningful representations of the information present in the input data, exploiting the structure and architecture of the neural network to learn models and relationships.
In summary, sensory neurons in artificial intelligence play a key role in processing initial information from sensors or input data, allowing the neural network to learn and interpret complex information in its subsequent layers.
chatbot
A chatbot, deriving from the combination of the words "chat" (conversation) and "robot", represents a computer application designed to simulate human interactions through a conversation. By exploiting artificial intelligence, the chatbot is able to interpret and respond consistently to user messages, providing an interactive experience. These virtual assistants can be integrated across different platforms, including websites, messaging apps, social media, and voice interfaces. Their applications are diverse and include automated customer support, information provision, reservation management and much more.
There are mainly two types of chatbots:
Rule-Based: These chatbots follow a predefined set of programming rules and respond according to predetermined patterns. Their interaction is limited to the logic established during programming.
With Artificial Intelligence (AI): These chatbots use machine learning and natural language processing (NLP) algorithms to understand user messages. They are able to learn from past experiences, improve over time, and deal more adaptably with complex interactions.
Chatbots have become increasingly popular, playing a vital role in areas such as customer service, business applications and online platforms. Their presence aims to simplify user-machine interactions, improving accessibility and offering a more intuitive experience.
Subscribe to our newsletter!