Finally, in ChatGPT, machine learning has the public’s attention.
Since its public release in the latter part of last year, the free chatbot program, which can produce a wide range of text that is astonishingly human-like from basic prompts, has garnered a lot of attention (and significant investment). The “fastest-growing consumer application in history, according to Reuters, gained almost 100 million active monthly users in just eight weeks.
The future of journalism, academic assessment, internet marketing, and computer programming, among other professions, are already the subject of intense debate. Even the latest prospective “Google killer” has been spoken about it.
The potential to aid in healthcare is really intriguing, without getting into any of those controversial subjects.
Let’s define the technology first. Using the GPT-3 family of large language models from the corporation, the chatbot was created by the artificial general intelligence (AGI) research company OpenAI (LLMs). It exemplifies conversational “generative AI,” which is essentially machine learning algorithms trained on vast amounts of internet data to swiftly produce new content (in this case, text) with little input. From what it has “learned,” it can really make anything.
The buzz around ChatGPT is justified even though its output is by no means flawless. Natural language processing (NLP) technology has advanced significantly since the program’s inception, both in terms of sophistication and capability and in terms of how quickly it has evolved.
Consider the fact that the GPT-2 deep learning neural network’s final model, which was developed using 1.5 billion machine-learning parameters, was released by OpenAI in November 2019. In the middle of 2020, the beta version of GPT-3 was released. It was trained using more than 175 billion machine-learning parameters. Microsoft had granted a license for its use in goods till September 2020. Moreover, new Microsoft application capabilities were powered by GPT-3 by 2021. Microsoft has used GPT-3 this year to add “intelligent recap” capabilities to its premium Teams application, including “automatically generated meeting notes, recommended tasks, and personalized highlights,” and has just revealed that the most recent release of its Bing search engine will include ChatGPT-like capabilities.
And as a result, a lot of people are reconsidering what is made feasible by generative AI, how rapidly it can occur, and where it will have the largest impact. Healthcare is no exception.
Doctors and healthcare professionals might really need a break from the administrative and information overload. Generic AI might be able to assist.
Now think about the potential applications of such a sepsis-specific AI model.
Sepsis is described by Mayo Clinic as a potentially fatal illness that develops when the body’s response to an infection damages its own tissues. Doctors say that sepsis is “hard to notice and easy to cure in its early stages, but harder to treat by the time it becomes visible” and that it can be brought on by viruses, bacteria, parasites, and fungal diseases. A perplexing “constellation of symptoms” may indicate sepsis. Moreover, it is the leading cause of fatalities in hospitals.
You could use something like BioGPT to:
• Examine voluminous biological literature for pertinent information on sepsis and extract it in order to find patterns and insights.
• Provide fresh research on sepsis in the biomedical literature, along with combinations of theories and hypotheses that might direct future investigations.
• Assist with sepsis diagnosis, treatment, and more swiftly locating intervention targets.
The existing healthcare coding and billing system may be affected by something like BioGPT, which relates to a totally unrelated but no less problematic sector of healthcare. The cost of coding and billing in the United States is quite high and “far exceeds that in comparable countries.” With generative AI, we could:
• Effectively automating the process of coding medical operations and services to make time and resources available for other important duties.
• Recognizing and fixing coding errors, which are frequent in the existing system and reducing the risk of refused claims and other financial penalties.
• Determining new and developing trends in medical procedures and services to guide the creation of new CPT codes, enabling healthcare providers to more precisely reflect the shifting environment of medical treatments and services.
The application of generative AI has the ability to significantly influence clinical comprehension, diagnosis, and treatment of difficult medical diseases as well as increase the precision, effectiveness, and efficiency of healthcare system operation. This technology may soon result in better patient outcomes, more efficient business operations, and lower costs for both patients and healthcare providers.
Though we’re not there yet, it’s not too early to begin considering how this technology can be of use.