Global interest in artificial intelligence (AI) has been on a steady upwards trajectory for the past decade. Since the onset of the COVID-19 pandemic, there has been a rapid acceleration of AI adoption–with 52% of respondents to a recent PwC survey saying their organization is accelerating its AI adoption plans. As artificial intelligence becomes mainstream, it is increasingly important that everyone, including those without a technical background, understand how the technology works and what it is capable of.
Although AI is often associated with a dystopian future where robots rule the world, this scenario is far from the current reality. Existing applications of AI in business involve highly specific use cases, and would be more accurately described as sophisticated pattern recognition instead of a replacement for human intelligence. This article provides an overview of AI, including different approaches to the technology, how it works, and the industries making use of it today.
Artificial Intelligence (AI) is a branch of computer science that deals with building smart machines that can perform tasks that typically require human brainpower. The concept of AI has been around for decades. However, recent advancements in machine learning (e.g., deep learning), rising computer power and declining compute costs, as well as the proliferation of difficult to process unstructured data has led to increased interest in, and adoption of, the technology.
One of the most authoritative AI textbooks, “Artificial Intelligence: A Modern Approach” by Peter Norvig and Stuart J. Russell, posits that there are four possible goals to pursue with artificial intelligence. Various approaches have been used to chase these goals throughout the history of artificial intelligence. Each has proven useful for advancing the field and producing valuable insights:
In order to claim a computer thinks like a human, we have to understand what human thinking is. Frameworks for how the human brain functions can be built through different methods, such as analyzing our own thoughts or conducting psychological experiments. After creating a cognitive model, it is possible to express the model as a computer program. Historically, people building systems that think like humans were concerned with comparing program reasoning to human reasoning (rather than focusing on the output regardless of how the machine got there).
The rational approach to AI emerged in the 1950s. The idea was that if human knowledge could be encoded into symbols that are linked together according to strict rules, then it would be possible for a machine to reason just as well as humans. This approach has not been very successful due to the sheer complexity of capturing any aspect of human reasoning with symbols.
Alan Turing proposed The Turing Test as a method of evaluating a machine’s ability to exhibit intelligent behavior on par with a human’s. Passing the test would involve a human operator asking the machine questions and being unable to tell if there is another person or a computer responding. According to Norvig and Russel, AI capable of passing The Turing Test must possess the following abilities:
This approach is focused on building AI that behaves as a rational agent. Building a system that acts rationally involves applying AI to real-world problems and allowing it to choose an action from a distinct set of choices. Models allow it to respond to unexpected situations and (ideally) make the best decision.
Modern applications of artificial intelligence, from automated customer service agents to AI-powered visual inspection systems, are solving real-world problems by building upon the various approaches taken to AI throughout history. This framework is also helpful for understanding what artificial intelligence is, what it's truly capable of, and the challenges to creating true human-level intelligence.
Artificial intelligence works by pairing massive volumes of data with quick processing that improves over time and enables intelligent automations. There are three main ways in which AI is usually implemented:
These various forms of implementing AI can be used for specific application scenarios. For instance, deep learning systems can be used for tasks like object classification or segmentation when building computer vision AI. Machine learning makes it possible for natural language processing systems to analyze texts and make inferences about intent and sentiment (i.e., positive or negative).
The specific applications of AI are vast given the potential for the technology to disrupt processes in virtually every industry. However, each approach relies on similar algorithms and principals and attempts to achieve the same goal: effectively recreating or simulating human intelligence.
Artificial intelligence has huge potential to disrupt the work all of us do every day. For this reason, it is often discussed in a future state. However, AI is already in use in a variety of real-world applications including:
Insurance companies are using AI to streamline claims processing, including automations for document processing, damage detection, and customer service. Automated vehicle inspection systems use computer vision to automatically detect automobile damage. No-code AI platforms make it possible for nontechnical business users to process and analyze unstructured data. This makes it possible to extract hidden insights buried in unstructured information and automate tasks that were previously manual.
The agriculture industry is leveraging AI to monitor crops and livestock, detect damage and diseases, and predict crop yields. Computer vision and deep learning techniques can be used to determine when crops are ripe, allowing farmers to harvest at optimal times. AI-powered solutions for agriculture can also automate crop health monitoring, making it possible to detect pests and diseases before they cause major problems.
Retailers are using AI to make personalized product recommendations, optimize inventory management, automate product listing quality control, and much more. According to a recent KPMG study, AI accelerated the disruption of the retail industry by five years in just 12 months. According to the same study, retail business leaders expect that in the next two years, AI will have its biggest impact on the industry in customer intelligence, inventory management, and chatbots for customer service.
At super.AI our mission is to make artificial intelligence more accessible, and to automate repetitive tasks so that people can focus on the work that matters. We take this approach in everything we do and strive to create useful resources that empower people to learn about and leverage AI. For more information or help getting started on your AI journey, check out the following resources: