Google’s new innovation in artificial intelligence (AI) has medical overtones. Within the advances are being made on the part of the company to try to slow down the progress of ‘ChatGPT’, something it tries to do with ‘Bard’. Under the name PalM 2, the company has created a new tool that demonstrates its capabilities to understand and generate languages. And now, to complement this progress, they have decided to make one that talks exclusively about medical ingredients.
One device, as reported by Google in recent days, has had to undergo an “approval score” with questions similar to those on the United States Medical Licensing Exam (USMLE). The model answered the broad and multiple choice questions accurately.
But beyond that, the importance rests on the disclosure data provided by the company by evaluating those responses. Expert medical level questions have been asked for the new edition. The result of the examination was 85 percent. This is an 18 percent improvement over the previous performance of Med-PaLM, which outperforms other similar AI models.
This model still has to work so that it can see the light of day in the real world. Regardless, these models were tested against 14 criteria including scientific feasibility, accuracy, medical consent, rationale, bias, and harms.
Maximize the security of responses
These were also evaluated by doctors and non-doctors and it was concluded that there were still gaps in meeting the standards of excellence and quality that Google seeks to achieve.
The intent is that this type of technology can aid in clinical decision making. However it can also be useful at training level for medical students and various branches of health to make their learning easier. Similarly, you will be able to analyze different data sets on a large scale to make it easier to identify certain disease patterns.
But, above all, it will also be useful for the patient as they will be able to get answers and provide information and support. However, this requires a high level of security so that they are not confused and do not cause confusion among people looking for that type of information. Rigor should be maximum to maintain the well-being of the advisor, as well as objectivity.
Google sticks its chest out with device for the toilet
“Med-PaLM 2 can help answer questions and summarize ideas from a variety of dense medical texts. We are working to add more capabilities to Med-PaLM 2, so that it can be used for mammogram-like medical synthesize information from images or help radiologists interpret results,” Google explained via its social network.
On the other hand, the company has analyzed the results of different regions. They claim that it is the first AI system to achieve a passing score of 72.3 per cent in the Multiple Choice Question Answer (MedMCQA) dataset containing questions from AIIMS and NEET medical exams.
Although it may contain statements, data or notes from health institutions or professionals, the information contained in Redaccion Médica is edited and prepared by journalists. We advise the reader to seek the advice of a health professional with any health questions.