Some thoughts about AI and Dentistry

About AI

 There has been a lot of “hype” attached to AI (Artificial Intelligence) and AI programs are disrupting the current paradigm of how many tasks are done.  A special edition of Time Magazine published this year flatly states that “AI may be the biggest game changer yet.¹”

AI programs are currently painting portraits, responding to emails, preparing tax returns, recording songs, writing publications, coding computer software, creating architectural blueprints, providing health advice, pricing medicine and houses, assembling cars, creating ads and determining which ads are seen on social media.²

AI proponents believe that AI platforms will not only change the way humans work and interact, but they will also lead to an explosion of creativity, scientific advancement, and unimaginable feats.  Some are forecasting that AI could add $15 Trillion to the global economy by 2030.  Thousands of researchers are endeavoring to expand AI’s capability.  It is estimated that AI computational power actually doubles every 6 to 10 months.

What is AI?

The original computing paradigm consists of coding a set of instructions for the computer to execute.  However, the AI computing paradigm or “generative” AI is completely different. It is based on the “neural network” of the human brain and is trained to “think.”  What AI does extremely well is identify patterns.  This AI paradigm has been ravenously adopted thanks to OpenAI’s ChatGPT, which “responds coherently–but not always accurately–to questions.”  Another popular AI platform is Dall-E, which allows the user to conjure up any desired image.  AI generated images have already flooded social media and one even won an art competition! Other AI platforms, such as GitHub’s Copilot, are expected to be widely adopted in the future. Copilot turns simple instructions into computer code.

Rumman Chowdhury, AI ethicist and CEO/cofounder of Humane Intelligence, wants people to understand that AI is not magic. “It’s simply math, put into code.”  She says that people think of programmers as “uniquely capable entities that are smarter than everybody else.”  There has been a mystique enveloping AI technology so that people think AI is better at making decisions than any expert of a particular field. Rumman emphasizes that this idea is “wildly untrue³.”

Is AI being adopted too fast?

“Even though in the long term…all of the hype is warranted, the short-term hype is pretty disconnected from the current reality of the technology,” says Sam Altman, CEO of OpenAI. Generative AI is less than perfect and it can produce errors by returning misleading or inaccurate content. It sometimes conjures up erroneous facts out of thin air!

Companies like Alphabet and Microsoft are particularly interested in using AI for search engines. However, Margaret Mitchell, the chief ethics scientist at the AI development platform “Hugging Face” believes that the absolute worst usage of AI is with search engines, because it often comes up with the wrong answers.  “If humans come to rely on AIs for information,” she says, “it will be increasingly difficult to tell what is factual, what is an ad and what is completely made up⁴.”

It’s easy to see why Microsoft and Google are investing billions in AI. But as they rush to grab their share of the AI pie for future profits, research designed to keep these tools safe is being left on the back burner. Negative consequences, in fact, have already occurred. Companies like Stability AI are facing lawsuits from artists and copyright holders who raise objections to their work being used to train AI models without permission.⁴

There are much larger consequences that might arise in the future. There is concern that already powerful companies like Google and Microsoft will be able to monetize AI to become even wealthier and more powerful.  If they end up dominating the AI marketplace, they could control who might have access to AI platforms⁴.

Damien Williams, an expert on social justice, has concerns about AI returning biased information. He says that unconscious bias (aka implicit or systemic bias) is subtle and is found in all sorts of internet places like Twitter, Facebook, and email messages. He is concerned that the machines will adopt biases for people of color and disabilities. “If we build new technologies with social justice in mind and correct bias quickly, we will have a chance to build a better, more equitable world for everyone,” he maintains.⁵

Damien worries most about these developing issues:

1. Deep Fakes: It is going to get increasingly more difficult to identify whether an image or video is real or fake.
2. Voice Clones: Experts believe that voice clones will precipitate unprecedented financial fraud and elder abuse.
3. Hallucinations: Generative AI is known to conjure up its own facts out of thin air.  Tools like ChatGPT has returned fraudulent medical advice, false details about individuals, and citations of nonexistant articles
4. Invasion of Privacy: The use of AI often necessitates permission to access an individual’s data. There is a legitimate concern that AI might violate personal data privacy.

There is also a concern about deliberate bias by governmental or special interest agencies.  Ethan Huff, in her article on Natural News reports that ChatGPT has been programmed to combat “vaccine hesitancy.” Arguing that all vaccines are “safe and effective, “ChatGPT is pushing users to take whatever the government and media are pushing at any given time.” For those paying attention, Ethan notes, ChatGPT is “artificial fake intelligence” because “there is nothing intelligent about getting vaccinated these days.”  ChatGPT has also been programmed to respond to questions about STIs with “regurgitated propaganda from the US Centers for Disease Control and Prevention.”  “ChatGPT is quite useless,” says a commentator at The Defender, a publication for the defense of children’s health. “If it does not like the discussion, it just stops and wants to go on to a different subject.  Definitely a tool for the very gullible.⁶”

Many researchers are alarmed that openness and transparency in AI programs like ChatGPT might one day be discontinued. They believe openness and transparency provides significant benefits to society, such as advancing innovation, competition, and research; protecting civil and human rights; and ensuring safety and security.  Closed models make it much easier to conduct disinformation campaigns such as government-sponsored combat of vaccine hesitancy.  On March 25, 2024, 23 organizations and 24 individual academics sent a letter to Gina Raimondo, Secretary of the Department of Commerce, urging her to support openness and transparency in AI models.⁷

Experts such as Yoshua Bengio, Scientific Director of the Montreal Institute for Learning Algorithms worries that AI machines have the potentially do great harm. “I’m concerned that powerful tools can have negative uses and society is not ready to deal with that,” says Yoshua.  “Let’s slow down.  Let’s make sure we develop better guardrails⁸.”

In fact, there are some experts who take AI worry to an extreme. Emeritus Toronto Professor Geoffrey Hinton, for example, actually regrets helping to bring AI technology into existence.  He and others worry that AI systems will one day become “self-aware” and smarter than humans. Self-aware AI would have complete access to all data and would be capable of ideas like self-preservation.  It might become “an exact replica of human intelligence,” complete with human-like characteristics such as stubbornness (the Star Wars robot C3PO comes to mind). If self-aware AI machines develop their own goals, they could potentially wipe out humanity (the Computer “Hal” in the movie “2001” comes to mind).  Manisha Sahu, a writer and entrepreneur, reminds worriers in her blog, “What is artificial intelligence,⁹” that at present, self-aware AI is only a speculative possibility⁹.

Yann Lecun, Chief AI scientist at AI Meta believes that existential fears surrounding AI are completely preposterous.  “There’s no correlation between intelligent and wanting to take over,” he explains.   “Even within the human species it’s not the most intelligent among us who want to be the leaders. In fact, it’s quite the opposite⁸…”

These fears make it clear, however, that safeguards designed to prevent potential negative consequences are a must.  OpenAI CEO Sam Altman proposes that researchers on the frontier of AI research be strongly regulated.  “These systems are already quite powerful and will get tremendously more powerful…We have to come together as a global community before very powerful technologies pose substantial risks,” he explains.   Sam believes AI generated content must be identified as such and that independent audits for adhering to safety standards must be developed.  He emphasizes that such regulation is already quite “doable¹⁰.”  Unfortunately, only 80 to120 researchers are currently involved in ensuring that AIs line up with human values.  By contrast thousands of engineers are endeavoring to expand AI’s capabilities.

Will AI take away your job?

A more realistic concern is that many  jobs that humans now perform will undoubtedly be performed in the future by AI platforms.  If that is to be the case, what will humans do? Erik Brynjolfsson, senior fellow at the Stanford Institute for Human Centered AI, maintains that the AI revolution is not  about automating jobs to replace people. “By far the biggest benefit is having AI work with humans and have them be more productive,¹¹” he maintains.  AI platforms could perform more of the mundane tasks, allowing employees to accomplish more creative tasks with their time.  In fact, a recent McKinsey survey estimated that AI could automate work that currently occupies 70% of an employee’s time.

The best way to use generative AI tools, says Amanda Johnstone, CEO of Transhuman, a technology research and development laboratory, is to think about the end outcome of what you wish to accomplish and “reverse engineer.”

How to get started with AI

There is no question that some individuals are afraid of using the new AI tools, and some avoid them altogether, hoping they will go away. The AI revolution mirrors the PC revolution of the 80s and 90s in some respects.  Many practitioners of that era refused to embrace computing when it first became commonplace.  Just like the non-computer using luddites of the past, the AI non-adopters of today will eventually be left in the dust.  Tomas Chamorro-Premuzic, a professor of business psychology and author of I, Human: AI, Automation and the quest to Reclaim What Makes Us Human, believes that everyone will eventually use AI tools.  “If you’re the only person not even trying it out or using it,” he says, “you might actually suffer¹¹.”

Every new skill has a learning curve, and so does AI.  Why not embrace AI for personal growth and future enhancement?  There is a lot of training available on how to use AI.  Minnalearn has a free introductory course online on the basics of AI. (www.minnalearn.com). Ville Valtonen, CEO and co-founder of Minnalearn wants students to learn the basics of AI so that they have the confidence to adapt as the technology improves.  “The AI field moves so fast,” he explains. “We think it’s really useful that people learn the high level things first, and then they can apply that in the world. That’s something you can use for much longer than a specific tool.”  The course is designed for the 99% of people who cannot code but want to understand AI.  More than one million people have already signed up.

LinkedIn Learning, the educational branch of LinkedIn offers more than 100 courses covering AI—both theoretical and practical applications. These courses are broken down into “bite-sized” chapters that may be under five minutes in duration. Pricing is done by subscription or individually for each course.

Many universities, like the University of Pennsylvania and Harvard University also have reasonably priced courses in AI¹¹.

The Use of AI in Dentistry

Dentistry is not practiced in a vacuum, and it is very much affected by advancements in computer technology, as all fields are.  AI is currently being used by early adopters, but it is already clear that it will eventually be adopted by all dental professionals.  Currently, the most commonly used platform is the generative AI program ChatGPT.  This program has had a faster adoption than Instagram or TikTok¹². It is accessible to all as in a basic free version.  I asked ChatGPT how AI is being used in dentistry, and ChatGPT returned a fairly detailed answer.

There are several exciting areas where AI is currently used in Dentistry:

1. Administrative Tasks at the Front Office: “Any task that has to do with data processing and anything that’s routine, AI will excel at,” says Kelly Monahan, managing director of Upwork Research Institute.¹³ AI can automate various administrative tasks, saving time and efficiency. Erik Brynjolfsson, along with researchers Danielle Li and Linsey Raymond, conducted a study of 5000 customer support agents at a call center that was using generative AI.  The AI tools were found to boost worker’s productivity, reduce attrition, and greatly assist early-career workers.  ChatGPT says that by leveraging AI for administrative tasks, dental practices can reduce manual workloads, minimize errors, improve patient satisfaction, and focus more on delivering quality patient care.

Here’s what ChatGPT¹⁴ reported when asked what tasks AI can accomplish for the dental front office:

1. 1. Appointment Scheduling: AI-powered scheduling software can manage appointment booking, rescheduling, and cancellations. It can also send automated appointment reminders to patients via text or email.
   2. Patient Registration: AI chatbots or virtual assistants can handle patient registration processes, including gathering patient information, insurance details, and medical history.
   3. Insurance Verification: AI tools can streamline insurance verification processes by automatically verifying patient insurance coverage and eligibility.
   4. Billing and Payments: AI systems can generate and send electronic invoices, process payments, and manage billing inquiries. They can also assist in claims processing and follow-ups.
   5. Patient Communication: AI chatbots can handle patient inquiries, appointment confirmations, and follow-ups, providing 24/7 support and improving communication efficiency.
   6. Data Entry and Record Management: AI software can automate data entry tasks, update patient records, and ensure data accuracy and consistency across systems. 

2. Imaging and Image interpretation: Image interpretation has always involved a degree of subjectivity.  Dentists often do not agree when reading the same radiographs.  The Pearl Company, which specializes in reading and analyzing dental X-Rays, estimates from studies that 43% of caries in dental X-Rays is undiagnosed, 20% of decay diagnosed from X-Rays is incorrect, and between 24% and 39% of calculus, margin discrepancies and periapical radiolucencies are completely overlooked.

 AI has the power to see subtle differences that human eyes can easily miss.  In a greyscale 2-D radiograph, each pixel has an intensity (brightness) that represents the density of the object. The AI algorithm can learn to detect patterns through direct vision and through comparison with thousands of similar patient X-Rays to make accurate predictions (Machine Learning or “ML”).¹⁵   The Pearl Company’s research demonstrates that, with this ability, “computer vision is proficient and often superior” to human vision.¹⁶ Pearl scientists estimate that the use of AI in reading radiographs actually detects 37% more disease than individual dentists.¹⁷

 ChatGPT notes that Pearl specializes in dental image analysis providing insights into dental conditions and abnormalities.  Another program called Overjet (https://www.overjet.ai/about-overjet/) is focused on identifying dental issues “like cavities, fractures and bone loss.”  Both tools, says ChatGPT, “can be valuable for dentists to improve diagnostic accuracy and treatment planning.” There are other tools like CariVu and Denti.AI that also detect and diagnose dental disease through AI image analysis.

Endodontic imaging:  In the endodontic world, ChatGPT reports that AI is being used to analyze X-Rays and scans to identify anatomical abnormalities and assess root canal treatment outcomes. 

3. Diagnosis and Treatment Planning: AI is currently being used in medicine for diagnosis. Erik Topol, a practicing cardiologist at the Scripps Clinic and author of Deep Medicine:  How Artificial Intelligence can make Health Care Human Again, believes that Google Searches and WebMD are adequate for diagnosis and that the ChatGPT is “going to be a much more refined way of getting answers to questions.”  He cited a case where a patient went to 7 neurologists who diagnosed her with “Long Covid.”  Her sister entered her symptoms into ChatGPT and it’s diagnosis was “limbic encephalitis.”  She was then treated for limbic encephalitis and cured.  AI has a tremendous advantage over individual physicians—the ability to analyze massive amounts of data—far more than human experts could ever manage.  As a result, AI can see patterns that humans cannot.¹⁸

On the other hand, AI is far from 100% accurate.  Dr. John Halamka, an ER physician and authority on poisonous mushrooms and plants, noted that AI was not very accurate when looking at pieces of mushrooms.

 Periodontology:  AI is being investigated as a tool for diagnosing periodontitis and classifying types of periodontal diseases.  A periodontal disease diagnosis is currently based on evaluating pocket depths and gingival recession, with the Periodontal Screening Index (PSI) quantifying attachment loss.  Researchers Hao Ding and his colleagues at the University of Hong Kong noted that the clinical evaluation of periodontal disease has low reliability because it is subjectively based on the clinician’s experience.¹⁹

 Oral Pathology:  ChatGPT reports that there is ongoing AI research for the development of AI algorithms for early  detection of oral diseases, such as oral cancer. Pathologic diagnosis is definitively determined from tedious examination of stained biopsy specimens on glass slides under a microscope and with radiographs. Many benign conditions also mimic malignant ones in clinical appearance.  Hao Ding and his colleagues  report that only 20% of biopsies are actually malignant. AI can greatly assist in preventing the terrible consequences that can arise from misdiagnosis.²⁰

Orthodontics:  Orthodontic treatment planning is usually based on the experience and preference of the orthodontists.  Every patient is unique, so the treatment plan must be custom-tailored to each individual. ChatGPT reports that an AI program called Dentem uses AI to analyze patient data and provide personalized treatment recommendations.  Many variables, Hao Ding and his colleagues note,  have to be considered in order to properly diagnose malocclusions.²¹ AI is an ideal tool for solving orthodontic problems.  Hao Ding and his colleagues point out that “the skeletal patterns and anatomic landmarks in lateral cephalograms can be clearly seen with the aid of AI algorithms.”  AI models can also assess data from cephalometric images, CBCT scans and intraoral scans to determine requirements for ideal orthodontic treatment.  These models can also suggest the best sites to segment the alveolar bone in orthognathic cases.  Chat GPT reports that there are also AI-powered solutions to track the progress of orthodontic treatments and recommend adjustments as needed.

AI models can be used to predict growth and development patterns in humans, and these predictions can have significant implications for dental treatment planning, such as when to place dental implants. AI algorithms can analyze dental and facial images to predict how a patient’s teeth and jaw structures will change over time.  AI models can also evaluate factors such as skeletal maturity, dental eruption patterns, and facial proportions to predict optimal timing for such orthodontic interventions as braces or aligners. “By leveraging AI-driven predictive modeling, dentists can optimize treatment outcomes and reduce the need for complex interventions in the future,” says ChatGPT.

Pediatric Dentistry:  AI can aid in predicting the eruption sequence of primary and permanent teeth, guiding pediatric dentists to better manage shifting dentitions and to address developmental issues such as overcrowding or early tooth loss.

4. Design and Fabrication of Dental Prostheses: AI is focused on improving the accuracy and efficiency of CAD/CAM systems for full and partial coverage restorations. Intraoral scanning has proven to be accurate for fabricating models, inlays and onlays, orthodontic aligners, bleaching trays, and mouthguards.

However, intraoral scanning is confined to tooth structure above the gingiva and is therefore not appropriate for creating crowns and bridges, even though it is commonly used for this purpose.  It has been reported that it is not yet possible to achieve individual custom digital designs for crown and bridgework.²² ChatGPT reports that researchers  are currently working on AI algorithms that can detect and correct errors in prosthetic designs, ensuring optimal fit and functionality.

The digital approach to full coverage restorations is not based on the sound principles of science and engineering that came from Dentistry’s “roots.” The incredible track record from Dentistry’s roots has been abandoned in favor of shortcuts and “workflow.”  Restorations made from digital intraoral scans are not accurate for crowns and bridges and, in fact, open the door to recurrent decay and loss of retention. More on this subject will be discussed in a future article.

However, desktop scans of analog models made with non-removable dies in the “analog” laboratory are quite accurate and milled restorations can be designed on the computer from these scans and then fabricated with high tech milling machines.  The Strategy Milling Company in Pittsburgh, PA²³ is milling highly accurate crowns and bridges from blocks of precious metal alloys that rival or exceed the accuracy of cast restorations.  The milling of denture bases has also been proven to be far more accurate than conventional heat and pressure approach, which almost always results in inaccuracies of fit, occlusion or tooth position.

AI directed digital algorithms can assist with smile design—visualizing and displaying the recommended treatment outcomes that result in satisfying outcomes.

Chat GPT reports that AI research is currently being conducted for the enhancement of “longevity and durability of dental protheses by predicting wear and tear patterns and recommending maintenance strategies.”

5. Patient Comfort and Education: ChatGPT notes that there are AI applications designed to improve the patient experience and treatment outcomes. Virtual reality tools, powered by AI, can help patients relax and feel more comfortable during dental procedures. AI-driven educational apps and games also promote oral hygiene habits and dental care awareness among young patients.”

In summary, AI is here to stay, and it is poised to revolutionize the entire profession.  The AI potential for improving patient care is limitless.  However, I believe that AI will never replace human dentists with robots. AI is a useful tool and nothing more. As Kelly Monahan, managing director of Upwork Research Institute explains, “Critical skills, strategic thinking, emotional intelligence, and creativity simply can’t be programmed. At its best, AI augments human potential—it’s not a substitute²⁴.”

AI has the potential to really enhance dental practice and open the door to endless fascination and learning.  Don’t be afraid to embrace it!

---

All the pictures were created with backgrounds from wallpapercave.com.

¹Editor, Time Magazine Special Edition; Spring 2024; p. 6.

²Chow, Andrew; Perrigo, Billy (with reporting from Leslie Dickstein and Mariah Espada); Time Magazine Special Edition; Spring 2024; “The Arms Race is Changing Everything;” p. 9. (Updated excerpt from Time Magazine, Feb. 17, 2023).

³Perigo, Billy and Henshall, Will; “The Big Players;” Time Magazine Special Edition; Spring 2024; p. 21.

⁴Chow, Andrew; Perrigo, Billy (with reporting from Leslie Dickstein and Mariah Espada);  Time Magazine Special Edition; Spring 2024; “The Arms Race is Changing Everything;”; p.9-13. (Updated excerpt from Time Magazine, Feb. 17, 2023).

⁵ Chow, Andrew; Perrigo, Billy (with reporting from Leslie Dickstein and Mariah Espada);  Time Magazine Special Edition; Spring 2024; “Bias in the System” by Damien Williams p. 29.

⁶Ethan Huff; Natural News.Com; “Globalists Programming ChatGPT to “Reduce Vaccine Hesitancy by Parroting Health Narratives;” https://naturalnews.com/2024-04-21-chatgpt-reduce-vaccine-hesitancy-public-health-narratives.html; April 21, 2024.

⁷Bankston, Kevin (Senior Advisor on AI Governance, Center for Democracy and Technology), and Hodges, Jennifer (Head of US Public Policy and Government Relations, Mozilla); https://www.rstreet.org/outreach/coalition-letter-openness-and-transparency-in-ai-provide-significant-benefits-for-society

⁸Chow, Andrew and Henshall, Will; Time Magazine Special Edition; Spring 2024;  “Do we face an Existential Threat”; p. 24.

⁹ Sahu, Manisha; What is Artificial Intelligence? Types, Uses and How it Works; https://www.analyticssteps.com/blogs/what-artificial-intelligence-types-uses-and-how-it-works; June18, 2021.

¹⁰Felsenthal, Edward and Perrigo, Billy; Time Magazine Special Edition; Spring 2024; “Like the Star Trek Holodeck; p. 59-61.

¹¹Shah,Simmone (with additional reporting from Linda Marsa); Time Magazine Special Edition; Spring 2024; “How to Make AI Work for You”; p. 34-39. (updated excerpt from Time, Aug 9, 2023).

¹² Chow, Andrew; Perrigo, Billy (with reporting from Leslie Dickstein and Mariah Espada); Time Magazine Special Edition; Spring 2024; “The Arms Race is Changing Everything;” p. 9. (Updated excerpt from Time Magazine, Feb. 17, 2023).

¹³Shah,Simmone (with additional reporting from Linda Marsa); Time Magazine Special Edition; Spring 2024; “How to Make AI Work for You”; p. 34-39. (updated excerpt from Time, Aug 9, 2023.

¹⁴ChatGPT was asked how AI is being used in Dentistry.  This section includes ChatGPT responses to this question.

¹⁵Hao Ding, Jiamin Wu, Wuyuan Zhao, Jukka P. Matinlinna, Michael F. Burrow and James K. H. Tsoi (University of Hong Kong); “Artificial Intelligence in Dentistry—A Review;” Frontiers in Dental Medicine; February 20, 2023; p. 5. http://frontiersin.org/articles/10.3389/FDMED.2023.1085251/full

¹⁶ The Pearl Company; “Can a Computer Identify Carious Lesions in Dental X-Rays as Accurately as Humans?”  https://www.hellopearl.com/guides.

¹⁷ The Pearl Company; “The Future of Dentistry is Powered by AI;” https://www.hellopearl.com/guides.

¹⁸Weintraub, Pamela; Time Magazine Special Edition; Spring 2024: The Future of Medicine; p. 40-43.

¹⁹Hao Ding, Jiamin Wu, Wuyuan Zhao, Jukka P. Matinlinna, Michael F. Burrow and James K. H. Tsoi (University of Hong Kong); “Artificial Intelligence in Dentistry—A Review;” Frontiers in Dental Medicine; February 20, 2023; p. 5-6.

²⁰Hao Ding, Jiamin Wu, Wuyuan Zhao, Jukka P. Matinlinna, Michael F. Burrow and James K. H. Tsoi (University of Hong Kong); “Artificial Intelligence in Dentistry—A Review;” Frontiers in Dental Medicine; February 20, 2023; p. 8.

²¹Hao Ding, Jiamin Wu, Wuyuan Zhao, Jukka P. Matinlinna, Michael F. Burrow and James K. H. Tsoi (University of Hong Kong); “Artificial Intelligence in Dentistry—A Review;” Frontiers in Dental Medicine; February 20, 2023; p. 6-8.

²²Hao Ding, Jiamin Wu, Wuyuan Zhao, Jukka P. Matinlinna, Michael F. Burrow and James K. H. Tsoi (University of Hong Kong); “Artificial Intelligence in Dentistry—A Review;” Frontiers in Dental Medicine; February 20, 2023; p. 9.

²³Strategy Milling Company in Pittsburgh, PA is at the forefront of milling precious metals.  [60 Leetsdale Industrial Drive; Leetsdale, PA 15056; 724-266-3467; www.strategymilling.com]

²⁴ Shah,Simmone (with additional reporting from Linda Marsa); Time Magazine Special Edition; Spring 2024; “How to Make AI Work for You”; p. 39. (updated excerpt from Time, Aug 9, 2023).