Machine learning uses algorithms trained on data to produce adaptable models that can perform a variety of complex tasks. Most studies suggest that that if you compare a specialist in oral pathology or oral radiology to an AI system, they achieve almost the same levels of accuracy and precision. Both are between 92% to 98% accurate. Different Kinds of AI AI encompasses a group of technologies and can be defined as when “a machine or a computer can perform a task that would usually require human intelligence,” says Dr. Nguyen. “AI can learn, create new knowledge, and solve problems through machine learning.” Machine learning uses algorithms trained on data to produce adaptable models that can perform a variety of complex tasks. One application of machine learning that is common is facial recognition programs. Machine learning algorithms analyze and process facial features from images or videos, which allows the system to learn patterns and characteristics unique to each individual face. Dr. Nguyen says there are some misconceptions about the type of tasks that AI can do. “For example, ChatGPT is a language tool, that’s what it does. If you ask it to predict when Bitcoin is going to reach a million dollars, it can’t do that. It just focuses on language,” he says. AI in Dentistry “At this time, computer vision is the strongest application of AI in dentistry,” says Dr. Nguyen. Computer vision trains computers to interpret and understand the visual world. Using digital images and deep learning models, computers can accurately identify and classify certain objects. “We could use it in oral pathology or in radiology to diagnose and identify different structures in an X-ray,” Dr. Nguyen says. “Most studies suggest that that if you compare a specialist in oral pathology or oral radiology to an AI system, they achieve almost the same levels of accuracy and precision. Both are between 92% to 98% accurate. But it takes about 28–30 minutes for an expert to give a proper diagnosis, and it takes AI about 30–32 seconds.” In dentistry, the diagnosis of caries can be controversial and somewhat subjective. “We don’t always agree, one dentist with another,” says Dr. Nguyen. “An AI system could help identify zones, but it would still ultimately be the clinician who decides what to do. Monitor it? Treat it? AI could be a tool to help in the final clinical decision.” Dr. Nguyen says there is potential for AI to Deep learning is a subset of machine learning, which differentiates itself by using more data. If machine learning uses thousands of data points, deep learning uses millions. “Deep learning mimics the way the human brain learns,” says Dr. Nguyen. “It works through a network of neurons and one can give it a bunch of information that is not labeled, that is all mismatched, and a deep learning neural network will classify it on their own and find its own logic to learn new information or skills.” In neural networks, there are hidden layers between the input and output of the algorithm; these hidden layers are where the algorithm does its work and information is analyzed, processed, and transformed into an output. “Some neural networks can be many hidden layers deep,” says Dr. Nguyen. Hidden layers are what make the complexity of AI possible, and they create intricate algorithms that are not easily understandable by humans. Hidden layers make it difficult to understand how AI systems process data and generate predictions or decisions. “We only see the outputs,” says Dr. Nguyen. “But the inputs are huge. For example, for ChatGPT, the inputs encompass most of the entire Internet up until 2021. And the outputs are the conversation you can have with the application.” 23 Issue 6 | 2023 | Issues and People
RkJQdWJsaXNoZXIy OTE5MTI=