The Dartmouth Summer Research Project on Artificial Intelligence gave birth to AI as a research discipline during a two-month think-tank hosted in the summer of 1956. The private Ivy League college invited twenty of the brightest minds to an innovative workshop to address how machine learning could be applied to solving the kinds of problems previously reserved for human thinking.
Although organizers were disappointed in the time spent arguing over the future of electronic capacity and functionality, the emergence of scientific theories from the New Hampshire workshop would have an immediate impact on identifying universal applications for advanced computer technology. A decade later, Dendral was developed as the first problem-solving program used in organic chemistry.
In the early 1970s, MYCIN was developed at Stanford University as the first artificial intelligence system designed to help doctors diagnose and treat infections in humans. The time-shared computer system was used to diagnose patients based on symptoms and test results. The network between clinical and biomedical researchers allowed for the diagnosis of bacterial causes of blood infections.
Clinical and Pharmaceutical Applications
Two decades later, Cedars-Sinai Medical Center launched CorSage as the first clinical application of artificial intelligence used to help cardiologists identify heart patients that were more likely to have another coronary event. By the turn of the millennium, deep-learning increased AI efficiency and medical applications began to transform the healthcare industry by improving numerous patient outcomes.
By using algorithms to analyze data from large language models, pharmaceutical researchers could more accurately identify trends and patterns to help predict the effectiveness of potential drug candidates for targeted patient populations. Today, artificial intelligence engines are being used daily to tailor treatments to the needs of individual patients and ultimately improve their quality of life.
Despite the potential benefits of AI in drug discovery and patient diagnosis, one of the major challenges is the limitations caused by the lack of suitable data. Although massive amounts of medical information have been semantically indexed for use by artificial intelligence engines, in many cases the data is of low quality or inconsistent, which can affect the accuracy and reliability of the results.
ChatGPT Diagnosed Condition after Doctors Couldn’t
A recent AI article published in Entrepreneur by Madeline Garfinkle recounts a frustrated mother’s use of ChatGPT to diagnose her son’s rare medical condition after seventeen doctors and three years of agony. Her four-year old was experiencing extreme pain during the pandemic prompting his mother to begin the search for an accurate diagnosis from an array of medical specialists, but to no avail.
Since its controversial launch in November of last year, ChatGPT has often been criticized for its potential harm to humanity by raising real concerns of plagiarism and the spread of misinformation. But frustrated by failed attempts of going from one specialists to another, the mother was convinced that previous conclusions were specific to each doctor’s area of expertise and not to the bigger picture.
After learning that ChatGPT was released to the public (by OpenAI) to provide answers to a user’s questions, the mother began sharing information about her son’s symptoms and test results, including various MRI scans and the child’s structural behavior. Programmed with the collective data, the chatbot diagnosed her son’s condition as Tethered Cord Syndrome. A neurosurgeon then confirmed the rare disorder.
What is tethered cord syndrome?
According to Stanford Medicine Children’s Health, tethered cord syndrome is a rare neurological condition in which the spinal cord is attached to tissue surrounding the spine. The condition can result in nerve damage and severe pain as the tethered spinal cord cannot move fluidly to keep up with the child’s lengthening spine as they grow. The problem often presents with two or more symptoms at once.
Physicians at the Lucile Packard Children’s Hospital at Stanford University use a team of specialized providers, including neurosurgery, radiology, urology, orthopedics and neurology, to determine the best approach. The most common procedure involves making a small incision in the child’s back to carefully cut the bands of the affected tissue to release the spinal cord and eliminate dysfunction.
Although medical researchers and physicians worldwide are excited about the future of artificial intelligence in improving their diagnostic approach, treatment protocols, robotic applications, and the development of future pharmaceuticals, no AI-based approach should be considered a substitute for human interpretation. Moreover, AI is dependent on the availability of high-quality input data to support its large language model.
Promoting Responsible Medical Uses for AI
Since the inception of the basic concept of machine learning and applications for artificial intelligence engines over sixty years ago, academic think-tanks have been used to address the risks associated with AI-applications, especially in the medical field. So, despite the possible benefits of artificial intelligence in delivering personalized healthcare solutions, it can’t replace a doctor’s decision making process.
On September 13, 2023, the United States Senate invited the top tech leaders to meet in a closed-door session to build a foundation for a bipartisan policy for the ongoing development and use of artificial intelligence. The meetings focused on how the priorities and risks of AI should be regulated to minimize “civilizational risk” and its potential impact on industries and the public. The group included CEOs of Meta, Google, OpenAI, Nvidia and IBM. All agreed oversight is needed.
The creators of the world’s leading large language models warn that chatbots can produce flawed results and are currently subject to errors and bias. Since the accuracy of medical diagnoses and treatment plans are essential, data errors can be catastrophic. Therefore, it is critical that the medical and scientific communities ensure consistency and accuracy of data used to train the algorithms. Plus, protecting compliance of patient data is always crucial to ensure privacy.
A Dual Role Emerges for Medical Students
Publicly accessed chatbots are being trained on information provided by the masses, but without high-quality data input, artificial intelligence systems in the medical industry will not be able to meet the demands for patients as well as the future healthcare possibilities they face. That’s where today’s medical student and tomorrow’s doctors have a major role to play in furnishing knowledge for future AI applications.
“If you’re still thinking that being a practicing physician only requires polishing your memorization skills,” warns Dr. Hans Wolf, founder of WOLFPACC Physicians Achievement Concept Course, “you might be wrong. Even as a first-year medical student you should focus on the how and why of the basic sciences. Our innovative approach for preparing for the USMLE or COMLEX licensing exam integrates the five main organ systems through physiology; and not your memorization skills.”
The speed at which artificial intelligence can advance accuracy and efficiency for delivering the highest quality of healthcare will depend heavily on incorporating the hidden insights of medical professionals through scientific literature. Large volumes of up-to-date knowledge must be extracted, summarized, and integrated across different disciplines to feed even larger, more powerful language models to ensure a brighter future for medical care.
Dr. Hans Wolf devoted decades to developing WOLFPACC’s Methodology for helping medical students understand how to apply the basic sciences that they learned in medical school to the clinical task at hand. Find out how WOLFPACC can help you apply the knowledge you’ve learned to ensure your successful career in medicine.