In recent years, algorithms and robotics have been installed in our day to day, to the point that there are computer applications that make decisions about us without being aware of it and others that are used to control aspects of our life.We see it in facial recognition systems to monitor spaces and identify people, for example.
The common point that these technologies have is that they act and decide automatically, sometimes without human supervision, hence we have to set limits on what they can do and what does not.A deregulated application of artificial intelligence -based technologies (AI) could cause both companies and governments to put their interests to the ethical aspect, which guarantees respect for the rights of citizenship, their privacy or their presumption of innocence.
To address all these issues, the European Union (EU) has already presented a regulation proposal that all member countries will have to follow.In it it puts limits to some applications of this technology but also makes exceptions.In the damn twitchery we have analyzed with specialists in this field what the regulation says and we answer this great question: why do it need to regulate artificial intelligence?*
Dismissed teachers and outlines accused of fraud based on "totally opaque" algorithmic decisions
Some recent examples show that this regulation is increasingly important: "In the United States, teachers are saying goodbye to algorithmic decisions.In Great Britain, in the postal service, an anti -fraud algorithm erroneously identified a lot of plots that not only lost their work but were accused of stealing.It was shown that the algorithm worked fatal.They were condemned for theft.Very important impacts at the level of guarantees, "says Gemma Galdón, director of Consulting and Ethics Foundation and algorithms auditor.
In the end, algorithms are computer command that companies patent and in that they are shielded not to clarify their operation: claiming that doing it would harm their economic interests.For Galdón, in Europe there is a "total opacity" when knowing the real scope of the algorithms.The regulations that have been presented seeks precisely to do the most transparent algorithmic processes and prevent discrimination between concrete groups of the population.
A European regulation that collects the problems that are seen throughout the world
The first proposal for a large -scale regulation of AI has come from the hand of the European Union, but the truth is that in other regions of the world automated technologies are used to control and monitor citizens.The Socialist Eurodiput Ibán García del Blanco, in charge of leading the proposal of the European Parliament for an ethical regulation of AI, explains that what the EU wants is that this rule serves to mark red lines globally.
"You have to take into account the international aspect, since we will have to agree with the rest of the world on the basic aspects.We are not superpower to use, with very powerful armies or unanimous positions but we are a regulatory power that often drags the rest of the world.What we do is that it is the general rule in the future, because if you want to operate in the common market you will have to meet certain standards, "said the Eurodiputa.
Once the European Parliament and Commission have presented its final positions, the negotiation dance continues with the entrance of the European Council, of which the political leaders of the member countries are part.This debate between the three EU main institutions, known as trilus, will be extended during the next months to negotiate point by point the final text.García del Blanco expects the text to see the light "at the end of this year".
What will be prohibited with these regulations?
In the commission's proposal, there is directly spoken of “prohibited practices” of AI.Some examples: systems that use “subliminal techniques” that influence people's decisions, the exploitation of “vulnerabilities” of groups and the implementation by governments of social valuation systems of its citizens.
In damn.This is already talking about the letter signed by seventy academics, professionals and activists who claimed the Spanish government the immediate brake of facial recognition tools (who use AI to identify people in public and private spaces) until the regularization of them.This regulation prohibits these types of systems, unless they are justified by a security issue.For example?To detect possible terrorist attacks.
The question is, according to Ana Valdivia, artificial intelligence researcher in migratory processes in King's College London, that to date there is no record that such a program has identified a terrorist."This is a regulation with political consequences.We are investigating which private companies are developing those algorithms and how much money the EU is giving them.Also what kind of metric and precision they have ", summarizes.
Who to most affect this type of surveillance systems?Valdivia emphasizes migrants.According to the expert, the Dublin Convention, which regulates the asylum requests of the people who reach the EU Mediterranean coast, uses algorithms to register its fingerprint, control its passage through the European territory and facilitate deportations.The problem of these decisions taken by artificial intelligence is that, many times, "what the algorithm says goes to Mass, you cannot answer".
Spain is now a pioneer in "control" the algorithms that influence the workplace
With the arrival of the 'Rider Law' the door has been opened for workers to know the algorithms that influence and make decisions related to their job.From this point and how Spain has become a pioneer by approving this type of legislation we talk to the labor lawyer Íñigo Molina, who explained that this new standard is "a step forward" since it will help "control" some parameters that until nowThey were totally opaque for workers.
Molina considers that the entry into force of this modification of the Workers' Statute makes the unions access to the parameters of these algorithms "before there is a decision" that the company has to justify, such as a dismissal."Now workers can be prepared for it and this is important since algorithms have biases that can directly affect them," he adds.
Not everything goes for security and work: simple examples of how the algorithms strain it
Helena Matute and Acute Ujue are researchers in artificial psychology and intelligence and how biases influence people, especially those related to technology.They developed a study in which they showed that people tend to trust the recommendations made by a computer system when we talk about two very specific cases: the election of a couple (or a league) and political candidate.
"We have to do that study from the algorithm, to see how we can do better algorithms, that are as effective as possible, how we can verify that they are really working well, that they do not enter that feedback loop that is inadequate."We are giving them an authority, understanding them that they are neutral, very efficient and effective machines and that is why we are trusting them".
* We have updated this piece to clarify that it is not an already approved and in force regulation, but that it can introduce changes in the negotiation phases.
FIRST FOADPUBLICATION OF START: 05/17/2021