With strict regulation and strict penalties. If it were to be adopted, there would definitely be no Palantir European Union.
After several preliminary studies, the first draft of the European Union’s regulation of artificial intelligence was born. The draft text, which is expected to be published next week, is Politico Acquired (a PDF document Available by AINews he is).
According to the project and in order to prevent the development of a miserable system, the European Union will strictly limit the so-called use of high-risk AI applications. However, it is not clear how it intends to harmonize the planned regulations Ursula von der Leyen (In our introductory photo) led by the Commission with the technological sovereignty it has also proclaimed.
Supported, tolerated, forbidden?
The regulation was justified by drafting the draft, which states that some applications of artificial intelligence (AI) may pose a risk and threaten the core values that are protected by EU law (such as human life, health, and of course, the fundamentals). Freedoms, including basic freedoms). Right to privacy). As Politico says, regulation will prevent the European Union from developing a fully-fledged watchdog country like the one it was built in China years ago. At the same time, with the framework, I would like to promote the development and application of MI systems that meet the public interest and protect the fundamental rights of individuals. However, it is almost certain that if the text is adopted, there will be an inability in the European Union to build a company like Palantir.
According to the draft, an assessment board should be established at the level of member states to test and evaluate dangerous AI systems. Any development affecting a fundamental right in the European Union must be submitted to the Evaluation Board. If this is done, an interesting situation may arise, in which decisions will also have to be made about developments that are in an advanced state (self-driving cars, HR systems for employment, etc.). The role of the Board will be, among other things, to test MI systems before they are put on the market and to assess whether the system has adequate human oversight.
The boards’ performance will be indirectly standardized by the European Union: a European Council for Artificial Intelligence will be established with one representative for each member state, with a seat for data protection authorities of the European Union and the European Union. The council will oversee law enforcement and share best practices with councils in member states. This control structure has already been criticized by the European Organization for Digital Rights (EDRi) for Politico. Representative of the organization, Ella Jacoboska According to him, if the national authorities had discretion, it would open a series of loopholes and increase the gray area in the MI market, as can be seen in countless other areas where the organizational structure is similar.
The draft will impose severe penalties for violators of the regulation. Companies can be fined € 20 million, or 4 percent of their global sales, for developing and marketing blocked MI systems.
It would be prohibited to use MI systems for mass surveillance and tracking of individuals (basically, profiling would also be prohibited, as standardized processing and analysis of data from various sources would be blocked). Apps that map people’s social status rankings based on data or, for example, evaluate a person’s reliability or personality traits based on behavior patterns will also be blocked.
The introduction of camera facial recognition systems in public places will be subject to separate authorization and will generally be tightened for the use of biometric recognition. And people need to be notified separately when they interact with an MI based system. However, law enforcement organizations will be an exception in many respects: for example, facial recognition technology can be used in public places to prevent and detect serious crimes (search for terrorists, etc.) if its use is limited in terms of time and geography. (By the way, camera surveillance is becoming more prevalent around the world, and in this area Budapest also holds a prominent place in the European Union.)
Too little, too much
The draft basically brought no surprises. The European Union Agency for Fundamental Rights published the Institute for Information’s Ethical Framework last December, and in January the European Parliament voted in favor of the regulatory framework. This stipulated that MI could be used primarily and exclusively for purposes in the interest of humanity, to advance the common good. In use, human control is essential, that is, a person must be able to change or disable the operation of the algorithm if it performs its task with unpredictable results.
The January vote indeed (364 yes, 274 no, and 52 abstentions) indicated that the European Union is very divided on this issue. (For example, advocacy organizations recently called for a complete ban on the use of facial recognition systems in an open letter to the European Court of Justice.)
On the other hand, strict regulation can discourage innovation in the European Union. Stratops tend to go to the US or other countries that handle developments less strictly (there may be new opportunities for Britain here). He said it could also lead to political tensions between the United States and the European Union AINews. Washington is concerned about China’s development, so it asked the European Union to more flexible regulation of MI developments.
in March Eric Schmidt The head of NSCAI (National Security Committee for Artificial Intelligence), the former chief executive of Google, told Politico that he said the EU strategy will not lead to success. The federation is not a competitor in this area, so it should instead seek partnerships with the United States as well as the major US tech companies that dominate MI developments.