Artificial intelligence is one of the big topics of the future. But in addition to the great advantages that the technology promises, there are also potential risks that need to be kept in mind. A study looked at the advantages and disadvantages of artificial intelligence.
Chatbots, self-driving cars and networked machines in digital factories give an idea of what our future will look like. The use of artificial intelligence (AI) brings companies many advantages such as higher efficiency, less repetitive tasks and better customer service. However, if the technology falls into the wrong hands, the potential dangers could quickly outweigh the great benefits.
The ubiquitous networking increases the susceptibility to malicious cyber attacks or technical failures, which can lead to widespread failures and enormous financial losses. Companies are also confronted with new types of liability issues, as responsibility for actions is shifting from people to machines.
Artificial intelligence: study also takes into account the risks of the technology
In its new study “The Rise of Artificial Intelligence: Future Outlook and Emerging Risks”, Allianz Global Corporate & Specialty (AGCS) identifies the advantages of the increasing implementation of AI in business and society, including the insurance industry, but also points out the risks that may arise through the new technology. AI, also known as machine learning, is software that is able to think and decide like a human.
“Whether for business, politics, mobility, health, defense or the environment: AI brings many advantages, but also potential risks. We urgently need preventive measures to reduce risk in order to maximize the net benefit of mass deployment and reduce unintended side effects, ”explains Michael Bruch, Head of Emerging Trends at AGCS.
Even today, so-called “weak” AI applications can perform certain tasks such as recognizing texts. In the future, “strong” AI applications will even be able to control autonomous vehicles, make more accurate weather forecasts, diagnose diseases, conduct financial transactions or operate and monitor industrial machines. AI often occurs in conjunction with other new technologies, such as the Internet of Things, data analytics or blockchain. According to Accenture, AI could double the annual economic growth rate in twelve industrialized nations by 2035.
AI risks in cybersecurity and autonomous driving
However, these potential benefits come at a price. AI further increases the already growing risk of cyber risks, which according to the Allianz Risk Barometer 2018 already represent one of the greatest threats for companies: AI-supported software could on the one hand reduce the cyber risk for companies through better detection of attacks, but on the other hand similar attacks facilitate because planning and implementation become easier. The same hacker attack – or a programming error – could be replicated on numerous computers. This makes major global cyber incidents more likely, which Lloyds said could cause more than $ 50 billion in damage. A half-day failure of a cloud provider alone could cause damage of around 850 million US dollars (Cyence).
In the area of mobility, high expectations are placed on AI applications. Self-driving cars should eliminate human errors as the main cause ofaccidents, and AI in traffic control systems should better control the flow of traffic. However, it is still largely unclear who is liable in the event of accidents and which ethical principles autonomous vehicles should follow in dilemma situations when the health and life of other road users are at stake.
Five risk factors of artificial intelligence
To identify emerging AI risks, AGCS focuses on five areas: software availability, security, accountability, liability and ethics. “A responsible development and introduction of AI must take these factors into account,” demands Bruch. With regard to security, for example, the race for the rapid market launch of AI systems could lead to inadequate reviews and thus no safe and functional AI applications are guaranteed. This, in turn, could trigger more product defects and recalls.
As far as the question of liability is concerned, intelligent AI agents could in future relieve people of many decisions, but they could not be held legally responsible for them. As a rule, the manufacturer or software programmer of AI agents is liable for errors that cause damage to users.
For AI-based decisions that cannot be directly traced back to the design or production, but which are made by an AI agent based on his interpretation of reality, no one could be made explicitly liable under currently applicable law. “As the number of damage caused by AI increases, it can quickly become expensive and tedious to leave decision-making to the courts,” says Bruch. “We need expert bodies that clarify the legal obligation to enter and work out a liability framework that exposes designers, manufacturers or sellers of AI products to limited tort liability.”
The insurance industry plays an essential role when it comes to minimizing or hedging new risks through AI applications. Traditional covers need to be adapted to protect both individuals and businesses. Insurers need to be able to provide better solutions to threats such as cyberattacks, business disruptions, product recalls, and reputational damage. In addition, new liability insurance models are likely to prevail, making manufacturers and software providers more accountable and limiting the strict liability of consumers.
Artificial intelligence: insurers as early adopters of AI
Due to the large amounts of data and recurring processes, the insurance industry is one of the industries that rely heavily on machine learning. “AI has enormous potential for insurers. Initially, it will support the automation of processes in order to accelerate and improve policy issuance or claims processing for customers,”explains Bruch. AGCS is already using robots to enter data into systems and is testing “bots” that automatically classify and process incoming mail, e-mails or calls.
AI-supported data analyzes will also give insurers and their customers a much better understanding of risks so that they can be reduced more effectively or covered by new insurance solutions. Last but not least, AI will also change the relationship between insurers and their customers because it enables, for example, self-service offers or advice around the clock.