(ORDO NEWS) — A study conducted by Dr. Isaac Cohane, an information technology and medical specialist at Harvard, together with two colleagues, showed the possibilities of using artificial intelligence in the medical field using GPT-4.
The results of the experiment were presented in the book “The AI Revolution in Medicine”, written by Cary Goldberg and Peter Lee.
In an experiment conducted in March, GPT-4 answered 90% of the time on the Medical Licensing Exam correctly, which was better than ChatGPT, GPT-3 and GPT-3.5, and better than some doctors already licensed.
GPT-4 proved to be not only a fact-finder, but also a translator who can process medical documents in other languages and paraphrase them in a format that sixth graders can understand.
The AI also offered helpful advice to doctors on how to communicate with patients and how to behave at the bedside.
However, the researchers noticed that GPT-4 is pattern-limited and cannot engage in full causal reasoning.
However, when the system was given data from a real case, it did the job, diagnosing a rare disease at the specialist level with years of study and practice.
Dr. Cohane notes that the use of GPT-4 in medicine can be beneficial, but she is not always reliable in her answers at this time.
The book provides numerous examples of AI errors of various kinds, and the authors suggest conducting multiple sessions with GPT-4 to improve the reliability of its responses.
AI is not ethically responsible like a doctor or nurse and can make assumptions based on inaccurate data. However, it does not intend to deceive or mislead.
The pace of development of artificial intelligence in recent years has accelerated significantly.
The development of machine learning technologies and algorithms, as well as an increase in computing power, make it possible to create increasingly complex and efficient artificial intelligence systems.
For example, in 2012, a neural network called AlexNet was able to win the ImageNet image classification competition with more than twice the score of its closest competitor.
Over the years, there have been significant improvements in computer vision, natural language, robotics, and other areas.
It is also worth noting that the pace of development of artificial intelligence depends on many factors, including research funding, the level of technical equipment and human resources in this area.
However, in general, it can be said that the development of artificial intelligence is very fast, and we can expect even more amazing achievements in the coming years.
—
Online:
Contact us: [email protected]
Our Standards, Terms of Use: Standard Terms And Conditions.