04 May 2023: UPSC Exam Comprehensive News Analysis

C. GS 3 Related

Category: SCIENCE AND TECHNOLOGY

1. The EU’s Artificial Intelligence Act

Syllabus: Awareness in the fields of IT, Computers, and Robotics,

Prelims: About European Union’s Artificial Intelligence Act

Mains: Regulating Artificial Intelligence (AI) systems and tools

Context

The members of the European Parliament reached a consensus to draft a new version of the European Union’s Artificial Intelligence Act.

The need for regulating Artificial Intelligence

• As artificial intelligence (AI) are technologies now being used extensively, their algorithms are becoming more advanced and are capable of undertaking a wide range of complex tasks such as driving cars, voice assistance, detecting cancer, recommending music, etc. 
  • With such advancements, there has been an increase in the potential risks and uncertainties associated with them.
• Further, several AI tools are regarded as “black boxes” i.e. even the ones who designed the tools cannot actually explain what process takes place within the tools to generate a specific output. 
• There has been an increase in instances wherein complex and unexplainable AI tools such as AI-enabled facial recognition tools have led to wrongful arrests.
• More recently, there have been concerns over “Deepfakes” that are generated using complex large language models (LLMs) like Generative Pretrained Transformer-3 (GPT-3) and 4.
• Industry stakeholders including Twitter CEO Elon Musk and Apple co-founder Steve Wozniak recently urged the AI labs to stop the training of AI models more powerful than GPT-4 for six months in order to mitigate the potential risks to society and humanity. 
  • In their letter, the industry stakeholders stated that “powerful AI systems must be developed only once we are confident that their effects will be positive and their risks will be manageable”.

Artificial Intelligence Act

• The Artificial Intelligence Act or the AI Act was first drafted in 2021.
• The Act aims to ensure transparency, trust, and accountability to AI and put in place a framework to mitigate the risks associated with safety, health, fundamental rights, and democratic values of the European Union (EU). 
• The Act was further envisioned to address various ethical issues and implementation challenges in sectors such as healthcare, education, finance and energy. 
• The AI Act also aims to ensure that AI in Europe respects the 27-country bloc’s values and rules.
• The key objective of the AI Act was to establish a balance between the promotion of the uptake of AI and mitigating the risks associated with the use of technology.
• Similar to the EU’s General Data Protection Regulation (GDPR) of 2018 which was regarded as an industry leader in the global data protection regime, the AI law also aims to make Europe a global hub of excellence in the field of AI.

Salient features of the Draft AI Act

• The draft AI Act defines AI as “software that is developed with one or more of the techniques that can, for a given set of human-defined objectives, generate outputs such as content, predictions, recommendations, or decisions influencing the environments they interact with”. 
• It also identifies AI tools based on machine learning (ML), deep learning, knowledge, logic-based and statistical approaches.
• The Act employs an approach to categorise AI tech based on the level of risk associated with the health and safety or fundamental rights of an individual. The Act mentions four risk categories namely unacceptable, high, limited and minimal.
  • The Act bans the use of technologies that fall under the category of unacceptable risk with little exception. 
  • Such technologies include the use of real-time facial and biometric identification systems in public spaces, systems of social scoring of citizens by governments that result in “unjustified and detrimental treatment”, techniques that distort an individual’s behaviour and technologies that have the ability to exploit vulnerable classes of society.
  • AI systems that fall under the limited and minimal risk category are permitted to use with very few requirements like transparency obligations.
• The draft version of the Act also prescribes various pre-and post-market requirements for developers and users of systems such as biometric identification and categorisation tools, AI tools used in healthcare, education, law enforcement systems, and tools that provide access to crucial private and public services.
  • Since AI algorithms evolve over time, post-market monitoring obligations also play a significant role.
• The Act envisages creating a framework that sets various parameters and undertakes a strict review of high-risk AI systems which are called “conformity assessments”.

Recent proposals for the Draft AI Act

• The AI Act that was drafted two years back only included one mention of the term “chatbot”. However, the advancement in the use of general-purpose AI tools and chatbots such as the ChatGPT has mandated a need to update the rules in order to catch up with the significant rise in interest in generative AI tools.
• Legislators also have shifted their focus towards the use of copyrighted material by companies using generative AI tools such as ChatGPT or Midjourney image generator, because such tools have the ability to train themselves using the data available on the internet. 
  • Companies using such tools will be mandated to disclose the copyrighted material used to develop their systems. 
• Furthermore, discussions are underway on whether all forms of general-purpose artificial intelligence systems (GPAIS) be categorised as high-risk. 

AI regulations across the world

• The U.S. – At present the U.S. does not have a holistic AI regulation.
  • The government has however notified a blueprint for an AI Bill of Rights (AIBoR) which has been formulated by the White House Office of Science and Technology Policy (OSTP).
  • The AIBoR highlights the risks associated with the AI tools to economic and civil rights and mentions five principles to mitigate such risks.
  • The AIBoR blueprint employs a sector-specific approach to AI governance instead of a horizontal approach like that of the EU.
  • AIBoR is similar to a handbook and is not binding legislation.
• China – China has come up with the world’s first nationally binding regulations that focus on specific types of algorithms and AI. 
  • China has passed a law to regulate recommendation algorithms with a focus on how they disseminate information. 
  • China has also enacted legislation that targets deep synthesis technology like those employed in generating “Deepfakes”. 
  • To ensure transparency and analyse how the algorithms work, China’s AI regulation authority has also developed a database of algorithms where developers have to register their algorithms and provide information.
• India – India has undertaken steps to ensure that AI technologies are developed and used in a responsible and ethical manner, including the development of the Personal Data Protection Bill.
  • The NITI Aayog came up with the National Strategy on Artificial Intelligence (NSAI) in 2018 with an aim to establish the country as an “AI Garage”.
  • The Union IT and Telecom Minister recently said that the government is not considering bringing a law or regulating the growth of artificial intelligence in the country.

Also read – Regulating Artificial Intelligence: RSTV – Big Picture

Nut graf: The rapidly evolving pace of AI tools and advancements in general-purpose artificial intelligence systems (GPAIS) such as ChatGPT require policymakers across the globe to accelerate their efforts to develop a robust AI governance regime to mitigate and prevent the potential risks associated with such technologies.