New Wisdom Report
Editor: La Yan
Advertisement
[Introduction to New Wisdom]What concrete and visible help will AGI bring to us on a practical level? There must be major innovations in the medical field that we cannot escape.
Why do we need AGI?
Many people may not have thought about this issue carefully and only saw the results. Some people believe that scientific and technological progress should be promoted unconditionally. As for the reason, it may not have been thought deeply.
Perhaps in the end, we will only come to a conclusion that makes our lives more convenient. And in what aspects and what kind of convenience it provides, it may not be that perfect.
Advertisement
In fact, this idea cannot be wrong.
After all, some things are driven by motivation first and then gradually realized. Some things are done first and then we see what they can do for us.
Today we provide you with an entry point in a real-life context to see what AGI can help us.
The most comprehensive doctor
First let's introduce a person, Greg Brockman.
Friends who are familiar with the AI circle should be familiar with him. He is the co-founder + president of OpenAI.
From 2010 to 2015, he served as CTO at Stripe. Since then, he has been the president of OpenAI to this day.
But today I’m going to talk about his wife, Anna Brockman.
In 2019, the two of them got married. Behind their happy marriage was his wife's physical condition that cannot be ignored.
In a recent tweet, Greg wrote: “After five years of experiencing multi-system pain in my body, my wife was recently diagnosed with a condition called Hypermobility Ehlers-Danlos Syndrome. (hEDS) genetic disease.”
Be aware that hEDS is an inherited connective tissue disorder that causes joint hypermobility, joint instability, and chronic pain throughout the body.
hEDS is also accompanied by a variety of other symptoms and related conditions, affecting many different parts of the body.
Greg's wife has been an actress for almost six years and is currently a fitness coach. One can imagine how much pain this disease would cause her.
Judging from the introduction of this disease, this is a comprehensive disease. It involves many systems throughout the body, such as orthopedics, cardiology, neurology, gastroenterology, dermatology, etc.
Greg tweeted that the current medical system is set up for individual specialties. There are so many doctors to see for hEDS.
“In the past five years, we have seen more doctors and various specialists than Anna has in her entire life. Most doctors only focus on the areas they are familiar with and fail to integrate these fragmented information.”
Later, Anna's doctor, who specialized in allergy medicine, listened carefully to all of her symptoms and problems, piecing together the details about her medical condition.
Greg said that as human medicine advances, there seems to be a trend toward increasing physician depth at the expense of breadth. But for patients, what we need is sufficient breadth and sufficient depth, both of which are indispensable.
The most ideal situation is that in the future, we can make this comprehensive medical service pocket-friendly, just like an expert team composed of doctors from many departments, working together to protect our health.
And this is where AGI comes into play.
Greg finally tweeted that although there is still a long way to go in terms of technology, AGI has to learn how to combine it with the supervision of human experts in high-risk fields like medical care, and how to deploy it together, but the prospects are already It's becoming clearer.
By working together among technology developers, healthcare providers, governments and society, there is hope for a future where better healthcare is available to all family members.
Many netizens also commented on the post to express their opinions.
Bacarella said that if medical AI can be as smart as an average doctor, and at the same time as patient, focused, and knowledgeable as GPT4, then it is estimated that there will be big changes in the future.
Paul also believes that when AI develops to a certain stage, new medical diagnosis and treatment methods will be able to be widely used, making various medical technologies accessible to the general public.
And this may be the field where AI should be most involved.
Google's attempt
You know, Greg’s idea has been confirmed by Google before.
Researchers from many top universities and medical institutions such as Harvard University, Stanford University, Yale School of Medicine in the United States, and the University of Toronto in Canada jointly proposed a new medical artificial intelligence paradigm on Nature, namely “General Medical Artificial Intelligence”, which can Flexibly encode, integrate, and interpret multimodal data in medicine such as text, imaging, genomics, and more at scale.
Paper link: articles/s41586-023-05881-4
Google Research and Google DeepMind have also jointly published papers to implement and verify the concept of artificial intelligence in general medicine.
Paper link: pdf/ 2307.14334.pdf
The researchers first curated a new multi-modal biomedical benchmark dataset MultiMedBench, which contains more than 1 million samples and involves 14 tasks, such as medical question answering, breast and dermatology image interpretation, radiology report generation and summary, and genomics Variant identification. Then a new model Med-PaLM Multimodal (Med-PaLM M) was proposed, which verified the realizability of the general biomedical artificial intelligence system.
This is a large multimodal generative model that flexibly encodes and interprets biomedical data, including clinical language, imaging, and genomics data, using only a set of model weights. Med-PaLM M's performance is comparable to the state-of-the-art in all MultiMedBench tasks, and even surpasses dedicated SOTA models on some tasks.
The paper also reports that the model can generalize to new medical concepts and tasks under zero-shot learning, cross-task transfer learning, and emergent zero-shot medical reasoning capabilities.
The paper also further explored the capabilities and limitations of Med-PaLM M. The researchers compared chest X-ray reports generated by the model and those written by humans for radiologist evaluation. In 246 cases, clinicians believed that Med-PaLM M The reports were better than those written by radiologists in 40.5% of the sample, also indicating that Med-PaLM M has potential clinical utility.
To train and evaluate the ability of large models to perform a variety of clinically relevant tasks, Google researchers collected MultiMedBench, a multi-task, multi-modal general practice benchmark dataset.
The benchmark consists of 12 open source datasets and 14 independent tasks, containing more than 1 million samples, covering medical Q&A, radiology reports, pathology, dermatology, chest X-ray, mammography, and genomics. fields.
Not long after that, Google went on to launch Med-PaLM 2, the second generation product.
It is the successor of Med-PaLM and is more powerful than its predecessor, achieving 86.5% accuracy on USMLE-style questions, an improvement of 19%.
Med-PaLM 2 is trained on a massive dataset of medical text and code, including medical journals, clinical trials, and textbooks. This enables it to understand and generate medical language with high accuracy.
Not only that, Med-PaLM 2 can also make inferences and inferences based on medical knowledge.
According to expert analysis, Med-PaLM 2 has the potential to revolutionize health care in multiple aspects. For example:
・Improving diagnostic accuracy: Med-PaLM 2 can help doctors comprehensively consider all the patient's medical information, including symptoms, medical history, and examination results, to determine the correct diagnosis for the patient.
・Increase efficiency: Med-PaLM 2 can help doctors automate tasks such as summarizing medical records and finding relevant information from research papers. This frees up doctors to spend more time communicating with patients.
・Improve communication: Med-PaLM 2 can help doctors communicate complex medical information to patients in an easy-to-understand way. This can help patients make informed decisions about their treatment.
・Reduced costs: Med-PaLM 2 can reduce medical costs by automating tasks and improving efficiency.
Currently, Google's Med-PaLM 2 is still in development, but it has the potential to have a major impact on the healthcare industry.
However, when it comes to Google, I still have to mention the medical conversation AI-AMIE, which was just released two days ago, and it also directly passed the Turing test! ?
As of now, Google is actively testing it in an effort to make it more widely available in the future.
I wonder if products like Google's, as well as other medical AI and even medical AGI that may appear in the future, can solve Greg's problem.
References:
https://twitter.com/gdb/status/1744446603962765669