What is it
Artificial intelligence is a tool that allows you to solve problems by simulating human logic, generating new content.
There are two large subsets:
- Machine Learning is a system trained to learn automatically from experiences;
- The Deep Learning is a subset of machine learning that uses artificial neural networks to imitate the learning process of the human brain. The system is trained to learn on its own by recognizing patterns across multiple levels of processing.
In summary:
| Machine Learning | Deep learning |
|---|---|
| Subset of AI | Subset of Machine Learning |
| It can train on smaller data sets | It requires large amounts of data |
| It requires more human intervention to correct and learn | It learns on its own from the environment and past mistakes |
| Shorter training and lower accuracy | Longer training and greater precision |
| Create simple, linear correlations | Create nonlinear and complex correlations |
| Can run training on a CPU (central processing unit) | It requires a specialized GPU (graphics processing unit) for training |
| It works with unstructured data such as images, videos, and audio | |
| Ability to abstract and learn complex concepts, adapting to dynamic and complex environments |
Examples:
| Machine Learning | Deep learning |
|---|---|
| Supervised: classify emails as spam or non-spam based on labeled data sets, including both inputs and outputs | Social Media – analysis of large quantities of images in order to detect users with specific content (for example for facial recognition) or in sentiment analysis to understand what public opinion thinks of a product or service, for example by scanning online reviews |
| Unsupervised: grouping data based on similarities, even without knowing in advance which categories they represent. Learning from unlabeled data, which includes only inputs | Finance – Neural networks can predict company values, identify threats and develop trading strategies |
| Healthcare – helps understand patient behavior, facilitating diagnosis and treatment | |
| Cyber Security – Algorithms can detect and mitigate threats such as viruses and malware | |
| Digital Assistants – Natural language processing (NLP) allows Chatbots and digital assistants such as Siri, Google Assistant and Alexa to provide intelligent responses | |
| Texts – machine translation, text generation |
The advantages of Artificial Intelligence
What are the benefits?
Assistance
Digital assistance tools available 24×7 allow you to answer the most common questions, resolve issues or scale them to human agents, reducing repetitive tasks.
Reduction of errors
Correctly programmed algorithms allow you to increase accuracy and precision, for example through robotic surgery systems, improving safety and patient outcomes.
Zero risks
The use of robots eliminates risks for humans, for example for bomb defusing, interventions in the depths of the ocean, space travel or in fully automated dangerous production environments.
Repetitive tasks
Some repetitive and boring tasks can be automated, allowing people to focus on more complex or creative ones.
Impartial decisions
If the absence of biased opinions can be verified in the algorithm, the decision-making process can be more accurate, for example by selecting candidates based on knowledge and skills rather than demographic data.
What to pay attention to
Artificial intelligence can be useful for learning or solving problems, but there are some aspects that should be kept in mind because they could represent disabling elements or lead to a redefinition of its functioning.
Oligopoly
AI data and processing power are concentrated in a few private entities that are more influential than governments, academia, and public research
AI in the Public Interest: Confronting the Monopoly Threat
Intellectual property
there are now few large concentrations of knowledge “drained” from the internet, with doubts about the protection of intellectual property
Fake news
Only human intelligence and sensitivity can distinguish between real and false facts and there is a risk of the proliferation of distorted information (for example through Deepfake).
Organizations with resources can consider customizing generic models based on their specificities, but the problem of the accuracy and truthfulness of the available data remains
Prejudice and discrimination
Through predictive and generative statistical techniques, the results depend on the quality of the model and the source of the data. They may not be accurate or appropriate, if the model contains (in the algorithm) biased opinions (BIAS), generating prejudices and consequent discrimination
Data selection
Be careful when selecting the data used to “teach”, favoring small, specialized models.
It is important to involve human resources in the process to validate the results of AI models, before being published or used.
The risk may be to use inappropriate information, not contextualized to a specific case, obsolete and incorrect, generating responses that can be counterproductive
High costs
It is a resource-intensive technology that requires constantly updated hardware and software components, opening up reflections on environmental sustainability
Obsolete human work
It replaces some types of human work, at an exponential speed (probably) higher than that of the human being’s ability to acquire new skills and professionalism, resulting in the need to accompany change
Give meaning
It is not able to understand the meanings, as the results are generated by stochastic models, which follow causal, probabilistic laws
Ethics and emotions
Ethics and morals are human characteristics that cannot be incorporated into AI.
Feelings, team spirit for achieving goals cannot be replaced by computers
Rules and regulations
The above aspects need to be managed at a national and international level
Organizational Considerations and Aspects
The future of artificial intelligence will depend on the ability to enhance its advantages by canceling or mitigating possible threats to the well-being of the community.
Artificial intelligence should not be seen as an alternative to human intelligence, but as a complementary tool that can benefit the community, and which should certainly be used when dealing with a large amount of data, “Big Data”, as it is capable of performing processes that are impossible for a human being in a short time.
Being an instrument it must be kept under the control, review and “approval” of the human brain. If only because that of Homo Sapiens has an evolution of no less than 50,000 years and even today neuroscience has difficulty understanding how it works (How far neuroscience is from understanding brains).
Organizational Design
The adoption of artificial intelligence tools in an organization involves a rethinking of the activities carried out by human resources.
The classic organizational design is based on an organizational structure with defined, differentiated and integrated roles with operational mechanisms. We need to find the right person, able to conform to the overall design, the culture, the rules (even if it is important to know how to generate diversity to innovate).
The roles are usually declined into profiles, activities, responsibilities and are subject to periodic maintenance as the context changes.
Today, in light of sudden changes (not only due to technological revolutions), we need to be able to equip ourselves to respond to challenges by asking ourselves who or what is able to process the right questions and provide the answers: human resources through development of knowledge and skills.
It is necessary to restore centrality to the human resource by opening a constructive, motivating dialogue, rediscovering the sense of knowing, knowing and participating, also well aware of the limits of each individual intelligence (the human brain is not perfect). This awareness can build a collaborative dialogue with new technologies, not an antithetical one.
A challenge that can be an opportunity
An opportunity for human beings to elevate their skills using the tools offered by artificial intelligence.
It would mean, for example, combining a role-based organization with a level expressed by the knowledge and skills expressed, demonstrated by the uniqueness of each individual person, to find solutions, innovate, support others. Remuneration should also be based on this.
Knowledge, skills, abilities and ideas of the future are the fruit of the unique history of each individual resource, who should have the possibility of making them explicit and developing them within team, area, organisation. The organization that aims to develop a competitive advantage should provide the concretely suitable context (organized knowledge, opportunities for micro learning, advanced training, etc.) the “humus”, the tools, the shared mechanisms. A level of true widespread, flat, non-hierarchical intelligence, but also on multiple “neuronal” levels/areas, capable of making decisions on the degree of adoption of artificial intelligence tools that can be integrated into the knowledge base.
AI must remain a Tool for organizational action
A level added to the normal organizational flow in which it is not the organization that tells you what you have to do, but it is you who – in the light of your knowledge – proposes what to do.
