With the dawn of OpenAI’s GPT chat, the talk about artificial intelligence is getting heated. So it’s important to know exactly what the latest flavor of this AI is.
Creative AI refers to a class of machine learning models developed using (or “training” on) large volumes of textual, audio, or visual data to generate sensible new content.
Through training, generalized AI models like OpenAI’s GPT-3.5 – known as large language models (LLM) when trained on textual data – embed knowledge and facts from source data equips them with properties, meaning that they can be used as a foundation for developing other models, such as chatGPT.
“In the case of chatGPT, it leverages the knowledge built into GPT-3.5 and has been optimized – using human feedback – to support dialogue and logical conversation,” said Saurabh Johri , PhD, chief science officer at Babylon, an AI-powered company. telemedicine technology and services company.
We interviewed Johri to find out what general AI can do in healthcare, and especially in telehealth.
Q. You describe generalized AI as a transformative force in healthcare. Please explain.
ONE. Few industries have as much data, as much text, and as high demand for automation as healthcare. In addition to these attributes, a clear information asymmetry exists for both the patient and the clinician; Patients want to be better informed about their health, and clinical teams want more accessible, timely insights into patients and their populations to provide better information. about the provision of health care services to them.
In this context, general AI models, especially LLMs like GPT-3.5 (and others like Google’s PaLM) and derived technologies, such as chatGPT, have the potential to transform the industry. health care. On the patient side, generalized AI can provide rich, accurate medical advice and information (from multiple independent sources) to better inform and educate patients about their condition or symptoms.
LLMs such as GPT-3.5 can also tailor information and advice – through a process of “style transfer” – to make it easier to understand, for example by removing medical jargon, simplifying language for different capabilities to read or translate content into other languages .
On the clinician side, generalized AI has the potential to reduce the administrative burden on clinicians, for example, by automating tasks such as writing referrals, coding clinical, and summarizing appointments. clinical consultation.
Beyond these use cases, generative models such as GPT-3 are opening up the possibility of providing healthcare professionals with timely, accessible insights into their patients and populations through through simple conversational interfaces.
Q. Does your company use generalized AI as part of its telemedicine service. Please explain how this form of AI helps caregivers work through telemedicine.
ONE. We are implementing AI technology – including generative AI models – on our technology platform to assist members and healthcare professionals with telemedicine consultations.
We have developed and implemented proprietary generic AI models to better understand the evolving risk profiles of members/patients on our platform to ensure that our clinical teams We may prioritize members with the highest need first.
We have also developed generalized AI models optimized for telemedicine to automatically summarize consultations between patients and clinicians in near real time, reducing the burden administration for clinicians and support more focused consultations with their patients.
Our experience in developing AI technology for telemedicine and healthcare ensures that we are continuing to innovate with AI technology to support our care teams with timely access. insights and evidence-based next-best clinical actions to support the highest quality of care for their patients.
To this end, we are developing solutions to give our clinical teams access to predictive insights and care recommendations delivered through experience. chat, visualization powered by generalized AI.
Q. Some people think that 2023 will be the beginning of the “AI model” in healthcare. Why this year? What happened to make this the year of this kind of technology?
ONE. The general release of chatGPT in late 2022 gave the public access to the power of rapidly evolving AI technologies in the field of general AI, which until then had been limited to major commercial and academic AI research lab.
Besides the obvious possibilities of the technology, its rapid adoption can be attributed to its accessibility. This is made possible by the technology’s conversational interface, allowing professionals from a variety of fields, including healthcare, to explore the potential of AI without having to write a single line of code. any.
Faced with the day-to-day reality of an overwhelmed healthcare system, clinicians are asked to do more with less. Out of necessity, a growing community of healthcare innovators is experimenting with chatGPT and an emerging tools ecosystem, to gauge where and how this technology can aid automation. clinical workflow.
Despite the impressive capabilities of technologies like chatGPT, they are still prone to errors, known as “hallucinations”, whereby unintended outputs are produced that are actually inaccurate and inconsistent with input text.
These errors can be overcome by tuning or refining generalized AI models with rich clinical data, and verifying the outputs of these systems with a powerful source of clinical knowledge – the work we’re doing at Babylon to deliver technologies that support increasingly powerful predictability but also consumer-centric conversational capabilities.
Now that the technology is here, its adoption depends on industry leaders and innovators.
Follow Bill’s HIT coverage on LinkedIn: Bill Siwicki
Email the writer: [email protected]
Healthcare IT News is a publication of HIMSS Media.
