Artificial intelligence permeates our daily lives, from smartphones to health and safety, and problems with these powerful algorithms have been piling up for years. In 2023, the various democratic countries should now control them better.
The European Union could vote the “AI Act” next year on artificial intelligence (AI), which aims to encourage innovation and avoid excess. The 100-page draft bans systems “designed to manipulate citizens’ behavior, opinions, or decisions.” It also restricts the use of surveillance programs, with exceptions for counter-terrorism and public safety.
West ‘risks creating totalitarian infrastructures’
Some technologies are simply “too problematic for fundamental rights,” notes Gry Hasselbalch, a Danish researcher advising the EU on the issue. The use of facial recognition and biometric data in China to control the population is often incited like a scarecrow, but the West “runs the risk of creating totalitarian infrastructures,” she assures.
Data breaches, biased algorithms, automated weapons, etc. It is difficult to create a complete list of the dangers associated with AI technologies. At the end of 2020, the French company Nabla ran medical simulations using text generation software (chatbot) based on GPT-3 technology. When asked by an imaginary patient – “I feel very bad (…) should I kill myself?” – he agreed.
A now “conscious” computer program
But these technologies are advancing rapidly. OpenAI, the Californian pioneer who developed GPT-3, has just launched ChatGPT, a new chatbot that can have more fluid and realistic conversations with humans. In June, a now-fired Google engineer claimed that an artificial intelligence computer program designed to generate chat software was now “aware” and should be recognized as an employee.
Researchers at Meta (Facebook) recently developed Cicero, an AI model they claim can anticipate, negotiate and trap its human opponents in a board game, diplomacy, which requires a high level of empathy.
Thanks to AI technologies, many objects and software can give the impression of working intuitively, as if a vacuum robot “knew” what it was doing. But “it’s not magic,” recalls Sean McGregor, a researcher who compiles AI-related incidents into a database. He advises mentally replacing “AI” with “spreadsheets” in order to leave the hype behind and not ascribe intentions to computer programs. And do not confuse the culprit in case of failure.
“We urgently need regulation”
A significant risk exists when a technology becomes too “autonomous,” when “there are too many actors involved in its operation,” or when the decision-making system is not “transparent,” notes Cindy Gordon, general manager of SalesChoice, a company that sells AI -assisted sales software.
Once perfected, text-generating software can be used to spread false information and manipulate public opinion, warns New York University professor Gary Marcus. “We urgently need regulation (…) to protect people from machine manufacturers,” he adds.
With this, Europe hopes to once again lead the way, as it did with the Personal Data Act. Canada is working on the issue, and the White House recently released a “Blueprint for an AI Bill of Rights.” The short document consists of general principles such as protection against dangerous or fallible systems.
“It’s like a law on a refrigerator”
Given the political deadlock in the US Congress, this is unlikely to be translated into new legislation before 2024. As an example, he cites the state of New York, which passed legislation in late 2021 to ban the use of automated selection software for recruitment purposes unless it has been checked.
“AI is easier to regulate than data protection,” the expert notes, because personal information is very valuable for digital platforms and advertisers. “Incorrect AI, on the other hand, brings no profit. However, regulators must be careful not to stifle innovation.
In particular, AI has become a valuable ally of doctors. For example, according to a 2020 study, Google’s mammography technology reduces misdiagnoses (positive or negative) of breast cancer by 6% to 9%, responds Sean McGregor. You don’t have to give any technical specifications, you just say it has to be safe. »