CATEGORY 102 AUSTRALASIAN DENTIST AI is all around us As technology continues to advance at an unprecedented rate, one of the most exciting developments in recent years has been the rise of artificial intelligence (AI). With its ability to analyze vast amounts of data, learn from patterns, and make datadriven de- cisions, AI is being adopted at an astonishing pace across a wide range of industries and applications. Take ChatGPT as an example. It took exactly five days to reach one million users and a few months to reach more than 100 million users1. For context, it took Netflix 3.5 years, Facebook 10 months, and Instagram 2.5 months. This incredibly fast adoption rate shows one thing: The world is ready for AI and there’s a race to implement it effectively to boost productivity and future-proof businesses for the long run. Despite the current surge in attention and the hype surrounding AI as a topic, the reality is that AI has been omnipresent in almost every aspect of our daily lives for the past decade and it has already played a crucial role in making our lives easier and safer. The facial recognition feature on our smartphones is not only convenient, but it also enhances security by utilizing biometric data to safeguard against unauthorized access and data theft2. Spotify’s “Discover Weekly” feature is a personalized playlist updated every Monday with new songs based on the listener’s history and behavior. In healthcare, AI has transformed diagnostics with 95% accuracy in detecting breast cancer3. It can help remotely detect signs of Atrial Fibrillation through our smart watches. We can now solve medical mysteries by tracing back an ultra-rare disease to a single gene through whole genome sequencing that can be done for less than a thousand dollars and in a matter of hours compared to when it was first done in 1990: it took 13 years to complete and cost around 3 billion dollars4. To narrow it down to dentistry, the literature shows a historically unprecedented boom in AI dental publications, with an average increase of 21.6% per year between 2011 and 20215 with the most impacted specialties being the following: Speciality Percentage Radiology 26.36% Orthodontics 18.31% General scope 17.10% Restorative 12.09% Surgery 11.87% Education 5.53% AI: a brief definition AI as a concept has been around for the past century, but the real revolution that made this vision come to life happened in the 1980s thanks to a computer scientist named Geoffrey Hinton, who took a step back from being an engineer and asked himself a highly relevant biological question: “What is intelligence?” His work was inspired by the structure and function of the human brain, as he has stated that the architecture of artificial neural networks is based on the structure of the brain’s neurons and synapses, and the learning algorithms used in neural networks are based on the way that the brain processes information (Fig 2). Today, AI is not one technology, it’s an umbrella term for a range of technologies and approaches (Fig 1) that often attempt to mimic human thought to solve complex tasks. How is AI developed? Developing an AI system relies on 4 crucial steps: Collecting data, labeling, learning, and validating & improving. See Table 1 below. How is AI reshaping the world of orthodontics? Harvesting data to improve clinical outcomes By Thomas Pellissard, COO & co-founder of DentalMonitoring Fig. 1: AI subsets AI is not one technology, it’s an umbrella term Fig. 2: Biological Neuron vs Artificial Neural Network. Source: Datacamp Building an AI system requires a very large amount of data to train it properly, but quantity isn’t everything — quality matters just as much. Collecting clean data means the system is trained on diverse data sets that are representative of the real world rather than being limited to one specific cohort, resulting in an unbiased system. Once sufficient data has been collected, there is a manual step in the process that requires human experts to identify attributes so that the algorithms can gain an accurate understanding of real-world conditions. Once the data has been collected and labeled, it’s added to the neural network so it can identify recurrent patterns. The model is rigorously tested, evaluated, and continuously updated and improved based on new data and feedback from users. Collecting thousands of photos of gingivitis cases. The photos collected should be high-quality and as diverse as possible in terms of population representation and degree of severity. The more diverse the dataset is the more representative of the real world it is. Dentists will manually label different characteristics of gingivitis on the photos and classify them into 2 categories: Gingivitis / No Gingivitis. After having analyzed enough photos labeled with “gingivitis”, the neural network will learn what characterizes gingivitis. Evaluate the AI’s analysis to determine if it is as accurate as a dentist’s diagnosis and correct any inaccuracies to continually improve the network. Collecting data Data labeling Learning Validating & improving Example of training the AI on detecting gingivitis Details Step Table 1: The steps for developing an AI system Thomas Pellissard FEATURE: ARTIFICIAL INTELLIGENCE
RkJQdWJsaXNoZXIy MTc3NDk3Mw==