IBM Watson can cause death

The IBM internal documents that have just been exposed show that many doctors used Watson to find that “the AI ​​gave “multiple unsafe, incorrect treatment opinions”, and even in extreme cases, Watson gave cancer patients with bleeding symptoms. A drug that causes bleeding is opened, and in severe cases, the patient may die.

Last year, a study published at the annual meeting of the American Society of Clinical Oncology showed that IBM Watson is very good at developing treatment plans for cancer patients. However, it turns out that artificial intelligence is still far from perfect.

Perhaps you still don't know, the obvious bug in the Watson system is not a sample in the laboratory. This Watson, which may kill people, has been used by more than 230 hospitals around the world, including 67 Chinese hospitals such as Peking University Cancer Hospital and Xuanwu Hospital. Distributed in 22 provinces.

image.png

A doctor at Jupiter Hospital in Florida told IBM executives, "In most cases we can't use it."

These documents came from a speech by IBM Watson's former deputy health director, Andrew Norden, before he left the company. Generally, only IBM management has permission to view it. A report from Norden in June and July 2017 shows that doctors using Watson for Oncology have pointed out that Watson often makes inaccurate medical advice.

A serious example of this is:

A 65-year-old man was diagnosed with lung cancer, and at the same time, the patient also appeared to have severe bleeding symptoms. Watson's advice to the patient is:

1. Receiving chemotherapy;
2. Use of the drug for the treatment of cancer, bevacizumab (trade name Avastin Avastin).

One of the side effects that bevacizumab may cause is that it is easy to cause bleeding. In the description of the drug by the American Association of Health Systems Pharmacists, this drug sometimes has serious or even fatal bleeding events, so the association recommends not taking it to patients with severe bleeding.

In addition to demonstrating customer dissatisfaction, they also revealed methods, processes, and underlying technical issues: Watson for Oncology should synthesize large amounts of data and present novel insights. But in fact, most of the data provided to it is hypothetical, not real patient data. This means that Watson's recommendations are based solely on the treatment preferences of a few doctors who provide data, rather than the actual insights gained from analyzing real-life cases.

An IBM spokesperson has said that Watson for Oncology "supports the care of more than 84,000 patients" and is still learning. But it is clear that what it learns is not based on real situations. According to media reports, there are at least 67 Chinese hospitals in 230 hospitals with this system.

Watson first entered the public eye in 2011, when it defeated humans in the television intelligence show "The Edge of Danger." The system has powerful natural language processing and database scanning capabilities, so you can understand the complex problems in the field and quickly give answers.

The effectiveness of Watson's intelligent supercomputer has been further questioned. IBM has invested billions of dollars in Watson's business, but "Doctor Watson" has yet to revolutionize the medical industry , and it is still difficult for IBM to make a profit.

Although the Memorial Sloan-Kettering Cancer Center believes that doctors are not stupid when they see a doctor, they will not fully listen to Watson, but the doctors' doubts and complaints will not be reduced.

First of all, this case shows that Watson's training is done through hypothetical cancer cases rather than real patient data. Previously, IBM executives have repeatedly told customers: "Watson for Oncology's advice is based on real patient data, and doctors around the world like it."

This made the outside world further question the validity of Watson. Jupiter Hospital in Florida is one of the angry and dissatisfied customers. In this IBM confidential document, a doctor at Jupiter Hospital said with anger: "This thing is just a fly! We bought Watson because its marketing is very powerful, so we feel that we can rely on it. It comes to achieve some vision. After the result is bought, this thing can't be done!"

After the incident was exposed, Sujal Shah, an oncologist at Jupiter Hospital, said that only when the doctors at Jupiter Hospital could not agree on two different treatment options, the doctors would look at the Watson station. Now 5:5, Watson, let's vote for it.

A previous STAT survey showed that Watson for Oncology did not meet IBM's expectations, and the exposure document showed that IBM executives knew that the treatment recommendations given by their products conflicted with national treatment guidelines, and doctors thought this The system does not work much.

Of course, IBM executives have always praised Watson.

The file that was exposed also showed that several IBM employees bluntly stated to Norden that the Watson product was "very limited".

In a statement to STAT, IBM said it has learned and improved Watson Health based on customer feedback, new scientific advances, new cancer treatment options, and more.

What is worrying is that this medical AI, which has been exposed to the wrong drug, has been put into use and has been used by many hospitals in the diagnosis and treatment of real patients.

An IBM spokesperson said that the Watson tumor system has been trained to help oncologists treat 13 cancers and is used by 230 hospitals around the world to "support more than 84,000 patients." Among the 230 hospitals, there are at least 67 Chinese hospitals (as of July 5, 2018).

Fu Gang, the chairman of IBM agent Baiyang, said in October last year: "At present, Watson for Oncology's services are mostly self-paying for patients, with an average of 5,000 yuan per person."

Participating in the training of Watson's Memorial Sloan-Kettering Cancer Center responded to the incident: "Although the Watson tumor system provides a safe treatment option, treatment decisions ultimately require the participation and clinical judgment of the attending physician, and no technology can replace the doctor's work. There is no technology that can replace the doctor's understanding of his patients."

Boiling Flask

Boiling Flask,Glass Boiling Flask,Round Bottom Boiling Flask,Borosilicate Glass Boiling Flask

Yancheng Rongtai Labware Co.,Ltd , https://www.shtestlab.com