https://www.bbc.co.uk/programmes/w172z0vv5fj788p
The EU is adopting a law that will regulate the development of artificial intelligence (AI) systems and protect citizens from them. On Wednesday, the European Parliament voted for the bill. This is not the end of the process: the three branches of EU power - the Commission, Parliament and Council - will have to agree on the final version of the law, its entry into force is expected no earlier than 2025.
The law divides AI systems into groups according to the degree of risk they pose to European citizens and consumers, and each group has its own restrictions and controls. For example, systems designed to influence the subconscious of people, as well as biometric recognition systems operating in real time, will be completely banned. Content-creating systems such as ChatGPT will, in particular, need to clearly label their products and not pass them off as the work of human beings.
“MPs want to ensure that AI systems are controlled by people, are safe, transparent, traceable, do not discriminate against users and contribute to the protection of the environment,” the European Parliament press service explained in May, during the drafting of the bill.
In the debate that took place on June 13, MEPs paid the most attention to points about facial recognition systems. The draft proposes to allow the use of such systems only "after the fact" (and not in real time), only to search for criminals, terrorists and missing people, and only by court order. A confident majority of MEPs is in favor of such a decision, but a small minority calls for a complete ban on facial recognition systems.
From this minority, Patrick Breyer, the only representative of the Pirate Party of Germany in the European Parliament, spoke in the debate. "We are opening a Pandora's box, we are creating a total police state, like in China," Brier said (quote from simultaneous translation). Leftists, greens and "pirates" in the European Parliament say that even the very limited use of AI facial recognition systems leads to the arbitrariness of intelligence agencies and the violation of the rights of citizens, as in China. The German "pirate" said that facial recognition systems are wrong in 99% of cases, and this leads to the detention of innocent people. "The court sanction demand proposed by hard-line conservatives is a pure formality, a smokescreen to cover mass surveillance [...] You can't normalize a culture of distrust and side with authoritarian regimes that use AI to suppress civil society," Brier said in front of debate.
His numerous opponents objected from the rostrum of the European Parliament that it is naive to demand a complete ban on recognition systems, that they are very useful in finding terrorists or, for example, missing children, and that the comparison with China is incorrect, since in Europe law enforcement agencies and courts really protect the rights of citizens.
According to Vice President of the European Commission Margrethe Vestager, who is in charge of antitrust policy, one of the most pressing problems associated with the development of artificial intelligence is discrimination. In an exclusive interview with the BBC, Vestager said that more relevant than the Terminator-style extinction of humanity is the potential for AI to increase bias or discrimination. This is facilitated by the vast amount of data that is being collected from the Internet and used to train models and tools, she says. Vestager told the BBC that AI regulation is needed to counter the biggest risks posed by the new technology. This is key, she says, when artificial intelligence is being used to make decisions that could affect a person's ability to earn a living, such as getting a mortgage.
"Probably [the risk of extinction] could exist, but in my opinion, the probability of this is quite small. I think that the risks of AI are more related to the fact that people will be discriminated against, they will not be accepted as they are," she said "If a bank is using AI to decide if you can get a mortgage, or if it's social services in your municipality, then you want to make sure you're not discriminated against because of your gender, color, or location."
As AI continues to evolve rapidly, new rules need to be pragmatic, Vestager said. "It's better to get, say, 80% now than 100% never, so let's start, and when we learn, we will return to this issue and adjust the rules," she said. In her opinion, the risk that AI could be used to influence the next election "definitely exists."
In the AI Law, which the central European authorities have been developing for two years - and recently amended in connection with the emergence of the ChatGPT-4 phenomenon - artificial intelligence systems are divided into four groups: "unacceptable risk" group, "high risk" group, "generating" systems and "limited risk" group.
The "unacceptable risk" group includes:
• systems designed to influence the subconscious of people
• social rating system that divides people into categories according to their behavior, socio-economic status and other personal characteristics
• real-time recognition systems
All of them will be banned in Europe, with an exception made for the use of recognition systems by law enforcement agencies after the fact and with the permission of the court.
Read also:
What should be the regulation of generative artificial intelligence?
Will Artificial Intelligence Lead to the Extinction of Humanity?