AI Act — EU attempts to tame the tech dragon.
EU’s Artificial Intelligence Act
Flashback to the year 1950, an eventful decade was turning in— The first credit card (made of cardboard) made its appearance, the first African — American Gwendolyn Brooks won Pulitzer prize for her poetry book ‘Annie Allen’, Disney released Cinderella which became its highest grossing movie of the year, ‘Peanuts’ the famous cartoon strip by Charles M. Schulz began as a daily strip in seven newspapers, U.S. declared its support to develop the Hydrogen bomb and Winston Churchill lost the general elections as the Labour party won but with a much reduced majority.
In the same year, Alan Turing released a path breaking and prophetic paper called “Computing Machinery and Intelligence”, popularly known as “Imitation Game”, in which he enquired — “Can machines think ?”. This was much before the term ‘Artificial Intelligence’ was coined in 1955.
Turing came up with an idea for a test that would measure a machine’s ability to exhibit intelligent behavior equivalent to, or indistinguishable from that of a human. This test became known as the Turing test.
Turing proposed that a human judge would engage in a conversation with a machine and a human, without knowing which was which. If the judge could not reliably distinguish the machine’s responses from the human’s, then the machine would be considered to have passed the Turing test and be considered as having human-like intelligence.
Years passed, and many attempts were made to create machines that could pass the Turing test. Some came close, but none were able to fully convince the judges that they were human. Nonetheless, the pursuit of artificial intelligence continued to drive progress in computer science and robotics.
And then in 1997, Deep Blue developed by IBM defeated Garry Kasparov the world chess champion.
In present day, we are seeing AI systems that appear truly intelligent, machines that can not only mimic human behavior, but also learn, reason, and solve complex problems in ways that were once thought impossible.
This raises an important question -
“Are machines above morality and more importantly above law? “
Regulators are now working on creating effective regulations, that will help to shape the future of AI in a way that is safe, ethical, and beneficial for all.
The EU AI Act is a proposed piece of legislation that seeks to regulate the development and deployment of artificial intelligence (AI) systems across the European Union.
EU AI Act: History and Timeline
2018 : EU Commission starts pilot project on ‘Explainable AI’.
2019 : EU Commission president Ursula Van Der Leyen talks about aim of making the EU a world-class hub for AI and ensuring that AI is human-centric and trustworthy through AI Ethics guidelines.
2020 : Consultation and evaluation of pilot phase of AI Ethics guidelines begins. Ursula Van Der Leyen releases the white book for artificial intelligence and data strategy and says EU wants to attract €20 billion per year through AI.
2021 : Draft “Artificial Intelligence Act” is unvieled. The act imposes checks on technology considered ‘high-risk’ and includes a ban on most surveillance and live facial scanning as well as AI systems to filter out school, job or credit scoring.
2022 : Launch of first AI Regulatory Sandbox.
EU AI Act: Definition of AI
To begin with a neutral definition of AI is established with an aim to cover all AI (symbolic AI, Machine Learning, and Hybrid systems) and include techniques that are not yet known or developed.
“a software that is developed with one or more of the techniques and approached listed in Annex I and can, for a given set of human-defined objectives, generate outputs such as content, predictions, recommendations, or decisions influencing the environments they interact with”
The quality of a robust law is that it is “Open” and “Abstract”, e.g. our Constitutions. Keeping this in mind, AI regulations are also being drafted in a manner that any future technology advancement and use cases can be brought into the regulatory framework easily.
EU AI Act: Summary of Ethical guidelines
As per the act, a trustworthy AI should be lawful, ethical and robust.
AI systems must satisfy the seven assessment guidelines — 1. Human agency and oversight 2. Technical robustness and safety 3. Privacy and data governance 4. Transparency 5. Diversity, non-discrimination and fairness 6. Societal and environmental well-being 7. Accountability
EU AI Act: A Risk Based Approach
The risk-based approach of the EU Commission’s AI Act draft refers to the way in which the regulation categorizes different types of AI systems according to their potential risks to human health, safety, and fundamental rights.
The AI Act classifies AI systems into four categories,
EU AI Act: What it could mean for Financial Services Industry
The provisions of the act are not expected to impact most of the core business of Banks, but only the AI systems which are high risk. Financial services will have obligations in their role as a software importer, provider, distributor and user.
Under the purview will be any system that uses machine learning or is logic-knowledge based or uses statistical methods.
Fines for violation of the rules can be up to 6% of global turnover, or 30 million euros for private entities.
CE marking indicates that a high-risk AI system is compliant with the regulation after undergoing a conformity assessment procedure.
Below table summarizes how the AI act will impact some of the use cases of AI in Financial Industry.
EU AI Act: The political-will backing this regulation
“EU’s vision for AI”
Through the ‘Horizon Europe’ and ‘Digital Europe’ programs, the Commission plans to invest €1 billion per year in AI and generate €20 billion revenue per year in future.
Paul Nemitz, a key contributor to GDPR and now actively involved in forming the EU AI Act, explains how unregulated AI can be a threat to Democracy.
“Democracy in danger”
“There has been reshaping of public space of discourse as algorithms are deciding what political news we see on the internet. There may be different variety of messages being sent from the same party which might not match their published manifesto. This is contrary to newspapers publishing the same news and point of view for everyone.”
EU Commission wants to propose a regulation of ‘Election Party Register ‘ that will provide a comparable overview of all messages a party pays for online discourse.
“Burden of Complexity”
Nemtiz adds further -
“Code is upgraded on daily basis which is difficult to keep check on. The big tech giants can take advantage of this for their business benefits while no-one else can transparently understand the changes.”
Another, political point of view behind the EU AI Act, is explained by Anu Bradford in her book -
“The Brussels Effect”
It states, in order to gain access to EU’s markets, companies must comply with EU regulations and standards. As a result, the EU is able to set global standards through its regulatory power.
With the AI Act, EU will have a starter advantage, as organizations would also push on having a consistent global framework to comply with and hence EU regulations will get world wide acceptance.
Conclusion
AI can be both powerful and unpredictable. Harnessing its power to improve our lives and solve our greatest challenges, requires us to understand how this technology works by improving our knowledge and skills. Creating frameworks to bring transparency and ensure use of AI for good, will be key in shaping the future. The Artificial Intelligence Act is an important step towards this direction as EU looks towards AI innovation for its economic growth. The draft still has loop holes however with support from governments, organizations and people the legal frameworks can be strengthened with time. In parallel we are seeing legal frameworks for AI systems management, being formulated by other strong democracies of the world like U.S., Britain, France, Canada, Japan and India.
For everyone's benefit we hope that the EU and rest of the world is able to tame and harness the power of the sharp-clawed beast called AI.
