What is machine learning and how does it work?
the term machine learning Refers to a subcategory of artificial intelligence consisting of mechanisms that allow smart machines to improve their capabilities and performance over time. The functionality of the machine learning system is based on: algorithm You can develop an inference model based on the provided dataset and adjust it over time according to your experience. In other words, these algorithms can adapt their structure and function based on the problem they solve. The more algorithms you solve, the more you learn. These properties make the algorithms used to create machine learning models increasingly accurate and precise, and theyautonomy” With behavior.
The starting point of every machine learning system is a set of data provided by the developer and consisting of: situation and solution variableeach represent a specific example of a problem the system must face and its solution.
who first showed interest in this field Alan TuringIn 1950, he submitted a proposal on the need for an algorithm that could analyze the first prototypes of machine learning. Since then, the technology has continued to improve and can now do incredible things.
different types of machine learning
Based on how the system gathers information and learns, we can define three different types of machine learning systems:
- Supervised machine learning. In this case, the computer is provided with data about the input, i.e. the problem presented to the computer, and information about the desired result, i.e. the solution to the problem. The goal of the system is to identify general rules that connect input data to output data so that these rules can be reused for similar tasks. In summary, examples of inputs and outputs are provided so that the system can learn the relationships between them and apply them to new cases.
- Unsupervised machine learning. Machine learning systems belonging to this category are provided with a dataset consisting only of input data and do not indicate the desired result. The goal of this kind of system is to “discover” hidden patterns and models. That is, identifying the logical structure of input data without prior labeling.
- reinforcement learning. In this case, the system analyzes the environment as well as past experience to improve the learning model. Reinforcement learning methods are characterized by reward/punishment mechanisms that control the functioning of the system. Indeed, while interacting with input data from both the dataset and the environment, the system is rewarded for achieving defined goals and penalized for undesirable behavior.
Also worth mentioning deep learning, a subcategory of machine learning that is very similar to reinforcement learning. Deep learning also autonomously creates inference models from datasets. However, unlike reinforcement learning, it does not proceed by trial and error. In particular, deep learning applies inference rules extracted from an ‘old’ dataset to a new dataset in order to identify relevant features. For this reason, deep learning is mainly used for recognition tasks such as speech recognition and image recognition.
Importance of machine learning
Regardless of the type of learning system, the data used by the learning system is training stage (datasets) play a fundamental role in the functioning of the system. This is because a large amount of information must be at your disposal to develop your algorithm. inference model We can solve your problem in an accurate and effective way. The results achievable with machine learning cannot be achieved with the human brain, which cannot process such a large amount of information. So it’s clear that this tool has great potential.
Consider how machine learning has transformed the financial system with forecasting and trading. cryptocurrency market: There are currently over 20,000 different cryptocurrencies, and AI can provide predictions for their respective data and graphics at every moment that exists in the metaverse. This would certainly be impossible for the human mind.
Technology for everyone and every day
This tool is often thought to apply only in highly technical fields such as science, medicine, space engineering, or other fields not commonly understood by non-technical people. This is a very common mistake because machine learning actually has many uses in everyday life. Of course, that includes using technology. A classic application of machine learning is home automation division: Tools that use speech recognition to learn new words and ways of speaking according to given voice commands are prevalent.
So for the first time in human history, we have a tool that allows us to track what millions of people do every day. Not long ago, only individuals at the top of society could boast of a biographer documenting their habits. , interests, behavior, but today we are all in the limelight (although often unwittingly). In fact, one of the most common uses of machine learning is profiling Individuals to enable businesses to create targeted advertisements. this is,”digital footprintIn other words, clusters of data generated as users use apps and navigate the internet.
Examples of the impact this profiling activity has on an individual’s life include the use of wearable devices such as: smart watchThrough the analysis of Collected dataIt can identify and track various activities performed by the individual wearing it during the day, such as tracking steps and heartbeat during the day. For example, by monitoring heart-to-beat variability, smartwatches can indicate when an individual is feeling stressed. A decrease in variability indicates a higher stress level and an increase indicates a lower stress level.
Additionally, smartwatches can track physical activity, such as your heart rate speeding up or slowing down while you sleep, and monitor a variety of other conditions. Personal profiling is also possible by using a car or other shared transportation, tracking movement via geolocation, and identifying frequent places and interests.
AI as a public sector booster
It’s worth emphasizing that artificial intelligence isn’t just for private companies. actual, Public institution Also, recently AI systemSome scholars refer to this “new age” as “digital state era” highlights the fundamental transformation of public activity through the introduction of new technologies.
For example, the European Commission’s Coordination Plan on Artificial Intelligence uses information and new technologies to innovate administration likewise. Following the initiative adopted by other European countries, Italy has initiated the following process. debureaucratization and Rationalization This creates a single, centralized big data structure to avoid duplication and significant loss of information and leverage the vast amount of information available to governments.
In fact, technology will affect many areas, from public services to the administration of justice.Regarding judicial administration, a particularly interesting application of machine learning is the so-called automatic decision making system In other words, an artificial intelligence system that replaces judges in deciding cases. Estonia is a pioneer in this field, and in 2019 he introduced the ‘X-road’ platform for the settlement of small bills under €7,000.00.
AI: Delicate Jigsaw Puzzles
While this is an interesting opportunity for businesses and public institutions, we must not forget that machine learning also poses serious challenges. riskespecially in relation to the processing of personal dataas shown in a recent episode Chat GPT and the Italian Supervisory Authority. The Italian authorities explained that Open AI did not provide information to the users and data subjects whose data was collected, and more importantly, from SA’s investigation, it “in turn collects and processes large amounts of personal data. There was no legal basis to support that.” It “trains” platform-dependent algorithms. For these reasons, Italian authorities have imposed immediate temporary restrictions on making Chat GPT available to Italian users and, consequently, on the processing of data by OpenAI.
It is not certain how supervisory authorities in other Member States will react to the adoption of this measure, but on the one hand, this perfectly underscores that the functionality of machine learning systems can be highly controversial. and also provides evidence of why Why both national and European institutions are drawing attention to this issue.For these reasons it is the responsibility of both industry and Public institution Share information about how the learning system works. This helps generate knowledge on the topic and allows people to trust artificial intelligence and use it in the best possible way.
The next step is the so-called artificial intelligence law.
