As digital technology spreads into everyday life, artificial intelligence has emerged to enable computer systems to perform tasks associated with human cognition and, in some cases, exceed human performance. Although definitions have varied, AI is now described as a technological direction using computers and machines to imitate human capabilities in problem-solving and decision-making. It analyses large datasets, identifies relationships and patterns, predicts events and responds to decisions. Tasks include visual perception, speech recognition, decision-making and language translation. AI encompasses machine learning, deep learning, natural language processing and robotics. Its three central capabilities are learning, reasoning and self-correction. The article distinguishes four types: reactive AI, limited-memory AI, theory-of-mind AI and self-aware AI.

Reactive AI is the most basic type and responds to current conditions. The source cites IBM’s Deep Blue supercomputer, described as developed in 1980, defeating world chess champion Garry Kasparov as an example. Reactive systems are programmed to produce predictable outputs from inputs. They respond in the same way to identical conditions and do not learn from past or future situations. They have no memory and perform only assigned tasks. The article gives spam filters, Netflix recommendation engines and chess-playing supercomputers as examples.

Limited-memory AI is more complex. It learns from data in a manner compared with connected neurons and can improve with experience. By analysing previous experience within limited memory, it guides future decisions, solves complex analytical tasks and makes predictions from historical data. Self-driving vehicles use this type, with algorithms trained on driving information. The article describes improved reaction times as supporting safe road use.

Theory-of-mind AI is a further frontier in which machines could make decisions with an understanding of human mental states. An important capability would be recognising emotions and adjusting behaviour accordingly. It has not been fully developed, although robots such as Kismet, introduced in 2000, and Sophia, introduced in 2016, have demonstrated some related features. A challenge is adapting quickly to changing human behaviour and emotion during communication. If achieved, robots could assist with daily tasks and act as companions. This type aims at emotionally intelligent behaviour when working with people. Researchers are making progress, but much work remains before systems can interact appropriately with each individual in context.

Self-aware AI would possess awareness of itself. Machines would understand not only others’ emotions and mental states but also their own. If developed, this would imply human-like awareness, intelligence, desires and feelings. No supporting hardware and algorithms have yet produced such a system. Researchers continue developing limited-memory and theory-of-mind AI.

AI is being applied in healthcare, finance, transport and the arts, with the potential to change many aspects of human life. Its development also raises ethical complexities and concerns. Algorithms trained on large amounts of data increasingly allow machines to do tasks previously performed only by people.

Machine learning, an important part of AI, develops algorithms that learn from data to make predictions and decisions. Deep learning is a subset using neural networks inspired by aspects of the brain. It produces highly accurate models for image and speech recognition, natural language processing and other applications. NLP develops algorithms to understand and interpret human language, supporting translation, chatbots and voice assistants such as Siri and Alexa.

AI is transforming healthcare through image analysis and disease detection. In finance, companies such as PayPal and Visa use it for fraud detection. Digital banking is also adopting AI beyond balance enquiries and transfers. The article cites Singapore’s DBS Bank and its Navi Planner, which recommends investments and insurance and projects retirement needs. Despite its benefits, AI affects employment, privacy and security. People worry about privacy intrusion, excessive influence, job losses as machines replace tasks, algorithmic bias and misuse. At the same time, the increased reliance on IT during COVID-19 highlights AI’s social benefits. International guidelines and frameworks should help ensure more beneficial and secure use.

This article encourages organisations and businesses to consider how to adopt innovation and AI, integrate new approaches and remain competitive in the future.

Nwe Ni Soe Yin

References

Report on the Study of New Digital Trends, Volume 1

https://bernardmarr.com/what-are-the-four-types-of-ai/