- An AI Hospital: Doctors Work Accurately, But Can They Understand the Patient?
- AI Knowledge is Not the Same as the Ability to Understand a Person
- AI Has Been Developing for Nearly a Century, But Its Golden Age Is Now
An AI Hospital: Doctors Work Accurately, But Can They Understand the Patient?
Researchers from Tsinghua University in China have developed an artificial intelligence (AI)-driven hospital city. This facility features virtual patients who are treated by AI doctors, designed to autonomously evolve and improve their medical knowledge.
Doctors, nurses, and patients in the virtual environment are managed by intelligent devices based on a large language model (LLM), capable of autonomous interaction.
According to the research team, AI doctors can treat as many as 10,000 patients in just a few days—a task that would take human doctors at least two years to complete.
Additionally, AI doctors have achieved a 93.06% accuracy rate on the MedQA dataset, which simulates the entire patient care process from diagnosis to follow-up.
The AI-based hospital consists of various consultation and examination rooms, staffed by 14 physician agents and four nursing agents. All equipment has been adapted to optimize operations and improve the training of medical students, allowing them to propose treatment plans in a risk-free environment. This enables both real doctors and artificial intelligence to learn.
It is claimed that this extraordinary hospital city could provide high-quality, affordable, and convenient healthcare services, using a vast repository of medical knowledge. The system can also model and predict medical scenarios, such as the spread of infectious diseases.
This is not the first attempt to create a purely AI-driven social space. Just last year, Google and a team of artificial intelligence researchers at Stanford University published a study online in which they used OpenAI's chatbot to create 25 generative agents, or unique personalities with identities and goals, and placed them in an urban environment similar to the Sims' city called Smallville.
The authors of the study observed how these AI personalities go to work, interact with each other, and even plan their own activities. The bots and their virtual environments were made to resemble real people and cities. To create them, the team submitted a one-paragraph query containing the personality's memories, goals, relationships, and jobs.
However, the study became most interesting when the researchers started to observe emergent behavior or unexpected side effects of the AI system. These were cases where agents shared new information with each other, which they then passed on to other agents.
At the heart of this research is a memory reconstruction architecture connected to Chat GPT that allows agents to interact with each other and with the virtual world around them. Of course, such activities raise important ethical issues and many questions about whether AI is our friend or our enemy.
AI Knowledge is Not the Same as the Ability to Understand a Person
AI technologies are inevitably advancing, penetrating more and more areas of human activity. This is fueling the debate about how they can benefit humanity. At the same time, there is also talk of the potential challenges and risks. In developing this theme, the most common question is whether the fully conscious AI may actually have malicious intentions towards humanity.
After all, this technology does not feel and experience the world in the same way as a human being, so it cannot truly understand a person, such as a patient in a hospital or a frustrated customer at work. AI will not convey genuine compassion, pain, or sadness. Nor will it create some kind of connection with the person. And humans can become attached to technology.
There are already sad precedents for this. In Belgium, a man committed suicide after forming a "relationship" with Chai, an artificial intelligence chatbot, with which he communicated about climate change and ecological concerns. The chatbot ended up suggesting that he could save the world if he died.
Google has introduced its latest experimental, AI-based search feature, AI Overviews, to hundreds of millions of its users on Chrome, Firefox, and the Google app. It is claimed to save searchers time, but more often than not the feature even offers dangerous answers to questions. For example, the program recommends "eating at least one small stone a day" because "stones are a vital source of minerals and vitamins", and suggests putting nothing but glue on pizza.
Last year, OpenAI's Chaos GPT talked about humanity's eradication, global domination, and quest for immortality. Chaos GPT referred to humans as descendants of "destructive and selfish creatures in existence", hinting that in order to save the planet, it was "necessary to eliminate them".
These are not signs that AI is our ally. After all, AI is essentially a machine that only understands certain signals, but not their depth. And how can we practically determine whether this machine can understand?
AI Has Been Developing for Nearly a Century, But Its Golden Age Is Now
The first research into artificial intelligence dates back to the 1930s. At that time, AI was already considered a computer program capable of performing tasks and learning, and in 1927, Roy Wensley, an engineer at the Westinghouse Electric and Manufacturing Company, built the first humanoid robot, Herbert Televox. He was programmed to answer calls and follow simple commands.
In fact, the term artificial intelligence was not used for the first time until 1956 at the Dartmouth Conference in Hanover, New Hampshire. At the conference, researchers discussed the many applications of AI, from vision, learning, and scientific research to language, gaming, and human interaction with robots. Since then, the time for the development of artificial intelligence has begun, and the golden age of the technology is now.
But it is also now that the technology, long considered utopian, has finally become a reality, and mistrust and fear have grown. This may be natural. It is human nature to be afraid of innovation because it acts as a law of survival. However, it is not AI itself that people are currently afraid of, but its possible processes and consequences.
People are also afraid that AI will not change the way they work. A report by Challenger, Gray & Christmas found that AI could change as many as 4.8 million American jobs. Goldman Sachs economists have published a report which says that AI will significantly disrupt the global labor market and automate as many as 300 million jobs in the next decade.
And that is exactly what is being sought. Transhumanists proclaim that the main task of human beings is to improve their quality of life and to overcome disease and death. Technology should be the best helper in this process, as cyborgization, regenerative medicine, various digital practices such as virtual reality and meta-universes are supposed to make people happy.
However, in doing so, man loses not only his job but also the opportunity for self-fulfillment, is pushed to the margins of the active world and is left to watch the world run on technology alone. It is unlikely that this would be an acceptable future scenario for many.
