Are we ready to use Artificial Intelligence (AI) as our next Doctor?
In the year 2371, in the Star Trek universe, the Doctor, an Emergency Medical Hologram, becomes the chief medical professional for the Star Ship Voyager crew when the spaceship gets lost in space and its human Doctor is killed in a conflict. The AI Doctor has a tough time adjusting to human patient biases in working with an AI and goes almost mad in a cathartic breakdown due to an ethical dilemma after having to decide to save one of two critically injured patients.
Now is the time for you to leap off your seats. In the last two years, we have made rapid strides toward such an AI due to lightening progress in Natural Language Processing (NLP), which focuses on understanding text and communicating in a human language. The astonishing success of OpenAI’s ChatGPT and similar models like Google’s BARD to communicate like humans using existing knowledge has opened newer and exciting possibilities.
ChatGPT and BARD are generative models. Unlike the previous generation of models, which focussed on making a prediction given a set of data, the generative models work by developing a better understanding of the text. They can use this understanding of the text to generate new text, catchy headers, viral tweets and even answer questions on complex subjects like healthcare. The models are further fine-tuned using a machine learning technique called reinforcement learning to emulate the responses of a human agent. The fine-tuning step makes the response of these models more human-like.
In Dec 2022, FlanPALM, another generative model from Google, passed the US Medical Licensing Exam. In January 2023, ChatGPT passed the US Medical Licensing Exam with more than fifty per cent marks. The exam is a prerequisite to be passed by any Doctor who wants to practice medicine in the USA. This is another hurdle crossed by the AI to become your Doctor, accompanying you on your next flight to Timbuktu or space. Human Biases and Legal challenges are now the more significant bottlenecks preventing this than technological limitations.
As humans, we are not kind to AI errors. In February of this year, stocks of both Microsoft and Google fell because of errors in responses by their AI during demos. Google’s ChatGPT incorrectly answered a question on the first photograph of an exoplanet. This drove a 9% drop in share prices of Google’s parent company, Alphabet. A similar 4% drop was observed in Microsoft share prices post-poor demo by ChatGPT integration in Microsoft’s web browser Bing.
Keeping all the challenges aside, generative models have immense potential to provide a 24-by-7 personal doctor to everyone, which can detect problems early by integration with users’ health devices such as smartwatches and address user queries. The continuing investment in AI will make these generative models better. However, social and political leaders must address the ethical and legal biases that impede the adoption of generative models as personal doctors. While we work on that, are you ready to visit the AI on your mobile device as your next Doctor?
ps: The article is my personal opinion and in no way reflects the opinion of my employer
