LONDON, June 21 (Reuters Breakingviews) – The key players in artificial intelligence (AI) are headed for a surveillance showdown. Lawmakers this month voted in favor of the European Union’s AI law in one of the first significant attempts to regulate the nascent sector. Other countries are also scrambling to make rules. Industry leaders like OpenAI boss Sam Altman warn that some approaches are too onerous, but the risks of pandering to special pleas outweigh the dangers of stifling new technology. .
AI is developing so quickly that there is not even a consensus on whether specific regulation is necessary. As venture capitalist Marc Andreessen points out, AI models are made up of code and algorithms like any other computer program. However, they are characterized by drawing their own conclusions rather than mechanically following instructions set by humans. In this way, AI enables photo generation tools to create photos that look like real humans. Additionally, the ability of AI models to extrapolate from data rapidly explains why nearly every industry is adopting this technology in hopes of increasing productivity.
These advantages are also why AI can cause big problems. By training a model on a given dataset, we can reproduce historical human biases, biases in mortgage approvals and job applications. AI’s ability to learn from freely available data also raises major questions about copyright law violations. And with technology likely to quickly take away so many jobs, it’s no surprise politicians are nervous.
Building guardrails is not easy. AI models can reproduce tasks performed by humans, making it difficult to identify which is which. The model’s ability to make apparently independent decisions also blurs the lines of responsibility, making it difficult for designers and public bodies to oversee it.
Despite these challenges, the government agreed on some general principles. A 2019 agreement brokered by the Organization for Economic Co-operation and Development stipulates that AI should be transparent, robust, accountable and secure. Beyond these generalities, however, there is a wide range of opinions about what counts as AI, what regulators should try to solve, and to what extent goals need to be enforced. There is a difference.
Policy makers are divided into various camps. At one end are the EU, China, and Canada, which are building new regulatory structures. At the other end of the spectrum are India and the United Kingdom, which said in their April white papers that, effectively, AI needs no special regulation beyond a similar set of principles articulated by the OECD. In the middle is the United States, where Congress continues to debate the need for targeted rules, while President Joe Biden is proposing an AI bill of rights. Such a divergence suggests the world is unlikely to set up a global AI regulator, and Altman pitched the idea to Congress on his May 16th.
The EU’s proposed legislation recommends classifying AI applications into four different buckets. A small number of “unacceptable risks” would ban the use of AI, such as real-time facial recognition for citizen surveillance. The majority are considered limited and low-risk and are subject to minimal oversight. AI systems that could be used to influence voters or election results, or systems used in social media platforms with more than 45 million users, are categorized as “high risk.” Altman’s complaint is that general-purpose AI systems, including OpenAI’s ChatGPT tool, underlie these applications, making them responsible for these risks.
He has a point. EU law mandates a “high risk” application to publish content generated by this technology and to publish a summary of the copyrighted data used to do so, and companies that make inappropriate disclosures will be It will impose fines equivalent to up to 7% of total revenue. This seems overkill for an application like ChatGPT, which is primarily used to summarize documents or assist in writing code. The stringent standards also discourage the development of general-purpose AI systems by small businesses and non-profits, restrict innovation, and challenge rivals like Microsoft’s (MSFT.O)-backed OpenAI and Google’s Alphabet (GOOGL.O). ) may be limited to industry giants such as As a result, the aggressive use of AI in fields such as drug development and semiconductor design may be curbed, and countries with lesser regulations may benefit from innovation.
But behind this new fangs, and perhaps game-changing technology, lies a familiar arm-wrestling race between regulators and big tech companies. Brussels and US watchdog groups are belatedly grappling with attempts to limit the power of giants like Alphabet, Microsoft and Facebook’s Metaplatforms (META.O). And the EU’s proposed regulation would effectively allow AI practitioners to mark their own homework. Given the speed of innovation and the risk of ChatGPT and other generative AI models creating more problematic applications, the threat of over-regulation is less daunting than the alternatives.
follow @karenkkwok on Twitter
(The author is a columnist for Reuters BREAKINGVIEWS. Opinions expressed are her own.)
(This story has been re-edited to add a hyphen in paragraph 5 and add a link to the penultimate paragraph)
contextual news
European Union lawmakers on June 14 agreed to changes to the draft artificial intelligence (AI) regulation, including a ban on the use of the technology in biometric surveillance and the use of AI-generated content in generative AI systems such as ChatGPT chatbots. required disclosure of
Among other changes, European Union lawmakers have asked companies using generating tools to disclose copyrighted material used to train their systems, as well as prohibiting “high-risk” applications. Companies that do are required to conduct a fundamental rights impact assessment and assess their environmental impact. It reveals that content is generated by AI, distinguishes so-called deepfake images from real images, and helps ensure protection against illegal content.
OpenAI CEO Sam Altman said during a speech in Paris that ChatGPT’s creators “want to make sure they can comply” in talks with European regulators, Bloomberg reported May 26. reported on the day.
On the same day, Altman said the company had no plans to leave Europe, rescinding a threat earlier in the week that it would leave Europe if it became difficult to comply with upcoming laws on artificial intelligence. . “We are excited to continue here and of course have no plans to withdraw,” he said on Twitter.
Editing: Peter Thal Larsen, George Hey, Katrina Hamlin
Our standards: Thomson Reuters Trust Principles.
Opinions expressed are those of the author. They do not reflect the views of Reuters News as committed to honesty, independence and freedom from bias based on trust principles.
