(Original title: Former AI "big brother", why IBM Watson is now considered a joke?)
Editor's note: As early as 2011, Watson won the quiz show Jeopardy! Since then, IBM has been actively promoting Watson. However, Watson has brought a lot of bad news recently. Watson worked with the Houston Anderson Cancer Research Center. The two parties split their way last year. IBM's revenue stopped growing, stock prices rose, analysts even began to question: Watson really bring more value? The well-known technology investor Chamath Palihapitiya said in an interview with CNBC in May this year: “Watson is a joke.â€
Paul Tang's wife had a knee replacement surgery at the hospital and he stayed with his wife. Every year, about 700,000 people in the United States do such surgery. When the surgeon arrived, Tang himself was the primary treatment physician. He asked the doctor when he expected to return to normal life and judge according to his experience. The surgeon has been vague and has no clear answer. Tang said: "I was really defeated. He didn't know it." Paul Tang quickly learned that most physicians basically don't know what the patient is like when he returns home and how he goes back to work. They cannot be patient. Life's assessment is important for patients.
Watson is a machine learning system and IBM will be betting on Watson. Watson can give physicians information, for example, telling the doctor how long it will take before the patient can walk or climb stairs without pain. It can also analyze images and tissue samples to develop the best treatment strategies for specific patients.
Since Watson won the quiz show Jeopardy!, health care has become a hot market for machine learning technology. Research firm CB Insights pointed out in the report that at least 106 startups have been operating since 2013 and they are still operating.
This year, Watson parted ways with the Houston Anderson Cancer Research Center. The criticisms for Watson (including the criticism of the Anderson Cancer Center) seem to have nothing to do with technology and do not mean that the technology is flawed. Only IBM is too optimistic about the current level of Watson, which only attracted criticism. In fact, when it comes to applying AI to health care, Watson Health is very likely to become a leader. If you don't have a big achievement right now, then the biggest obstacle to data is that IBM needs specific types of data to train the system. In a variety of environments, data supply is seriously insufficient and it is difficult to obtain. This problem is not just Watson's problem. It is the same for other systems. This is a problem for the entire health care machine learning field.
Lack of data and lack of data may slow down Watson's growth. Similarly, IBM's competitors will suffer. why? The best way to get the data is to work closely with large health care agencies, which are somewhat conservative with technology. Compared to startups, there is one thing that IBM does better: it has the trust of large institutional executives and IT managers. Although the Anderson project has some problems, IBM still has a great advantage. Many medical centers, health care management groups, and life science companies have all introduced Watson, and they all provide key data that lays the foundation for future medical models.
The schedule is impractical
Watson and the Anderson Cancer Center "split up" that IBM overtook Watson, resulting in trouble.
In 2012, Anderson Cancer Center and IBM began to collaborate. Their goal is to allow Watson to read data involving patient's symptoms, gene sequences, pathology reports, and then integrate information with physician reports and important journal articles to help doctors develop Diagnosis and treatment strategies. IBM and Anderson Cancer Center expect too much from technology. In 2013, IBM once remarked that Zhuang said "calculation of a new era is coming." This kind of assertion gives Forbes an illusion. It seems that Watson "can now deal with clinical trials." It can be used on patients in a few months. Now. In 2015, the Washington Post quoted an IBM Watson manager as saying that Watson is building a collective intelligence model, between machines and people. The Washington Post also stated that the goal of the machine system during training is to become a doctor's assistant and help them do what they could not do before.
In February of this year, the University of Texas (Anderson Cancer Center operated it) announced the end of the project, and the Cancer Center paid IBM $39 million in compensation. The original project contract was only US$2.4 million. After four years of research, the two sides did not develop a tool that can be used on patients. The technology can only be used in experimental tests and cannot be put into use. With regard to Watson, the Anderson Cancer Center declined to comment. It did not know how to manage the project internally and how to obtain funds. This may be the source of the problem.
It's not that IBM Watson has no problem. In fact, it has more problems than any other project.
What makes research frustrating? First of all, you should thoroughly understand the training methods of machine learning systems (such as Watson). Watson continues to fine-tune internal processing processes and learn from them so that the probability of getting the right answer when solving certain problems will increase. The correct answer must be known so that the system knows when it is right and when it is wrong. The more training problems the system handles, the better the results will be.
It is still relatively simple to train the system to identify malignant tumors in X-rays. If the puzzles being solved go far beyond the known categories of human beings, such as understanding the relationship between genetic mutation and disease, Watson will be powerless. It will encounter the problem of “chicken and eggâ€: the data has not been screened by experts and has not been effective. Organized, how to use such data training system? Thomas Fuchs, a computational pathologist at the Sloan-Kettering Cancer Center in New York, explained: "If you're training a self-driving car, anyone can label trees and signs so the system can learn how to identify it. But medicine is a special one. In the field, requiring specialist training for decades, labeling information correctly, and entering computers."
IBM hopes that Watson will contribute in some areas, and other companies' machine learning solutions are also intended. There are obstacles in all these areas. To train Watson to process huge amounts of data, pick out a few important information related to individual patients. First, people need to personally train and train with thousands of cases.
For example, in order for Watson to identify genes related to disease, it requires thousands of medical records. These patients suffer from specific diseases and their DNA has been analyzed. However, it is rather difficult to obtain data on the combination of "gene-patient records." In many cases, the data is not recorded in the correct format, or it does not exist at all, or the data comes from dozens of different systems and is difficult to process.
If you give better data to the clinician, you can improve the level of primary treatment. In the daily primary treatment process, when the problem is not serious, if the doctor misses the treatment opportunity and waits until the condition becomes serious, the patient goes to the emergency room or asks the expert to treat. At this time, the suffering will be greater, and the cost will also be greater. A substantial increase. Anil Jain, chief medical officer at IBM Watson Health, said: "One-third of the money spent on health is not needed." People think that machine learning has the opportunity to solve this problem.
When diagnosing a patient, in order for the doctor to come to a better conclusion, Watson needs to find a connection between the health records and the social determinants of health. These factors include: whether the patient is taking drugs, whether the diet is healthy, and whether the breathing air is fresh. Tang believes that today, there are few hospitals or medical practices that obtain reliable data from a large number of patients. This is partly because doctors are slow to accept modern data-driven medical practices. Manish Kohli, a physician and health care specialist at the Cleveland Clinic, said: "The health care industry is slow to accept technology and it is embarrassing."
If such data exists, IBM will generally pay for it. IBM acquired a number of companies, such as Truven Health Analytics, Explorys, and Phytel, which are companies that deal with big data. These data come from hospitals and patient groups. Although the cooperation with the Anderson Cancer Center was terminated, IBM has reached an important cooperation with other agencies to obtain more patient data.
Working with IBM, Atrius Health is a network of nearly 900 physicians, primarily primary care physicians, from the Boston area. The goal of the collaboration is to develop and test Watson-based system functions and extract key information for individual individuals from notes, records, and articles. Atrius Health chief medical officer Joe Kimura said: "For primary care physicians, extracting all relevant information is a heavy task." He also said that each visit to the data will increase, with such system data will increase significantly , do not need to submit in accordance with the standard format, the search is very convenient.
Also, many of the important notes in the patient's medical history are in the form of sentences that cannot be identified by traditional IT systems. Watson used natural language processing technology that was developed for Jeopardy! to extract meaning from sentences. Under ideal circumstances, the system can give advice to physicians, give patients better help, and save unnecessary care. Kimura said: "The patient's hips are hurt. We take care of it wholeheartedly. Why are we only concerned about this? Why can't we predict in advance and judge the risk of falling patients and prevent them from hurting the hips? We need to extend nursing care to the upper reaches."
Watson Health also works with the Central New York Care Collaborative, a state-funded agency that works with approximately 2,000 health care providers in six countries. The goal of the cooperation is clear: reduce the number of emergency and rehospitalizations by 25%. Sometimes the patient has been approved for discharge, and the result is returned to the hospital because of related problems. Cooperation has also brought a lot of patient data.
Get more data
There are other ways to obtain data. Google has a sister company that wants to extract data directly from patients. The company is Verily Life Sciences, which is Alphabet's health care division. The company is working with Duke University and Stanford University to develop a highly structured health database with 10,000 volunteers. The database contains outpatient information, as well as data from wearable health monitoring devices. As a result, the data will be greatly increased. It is only necessary to generate highly available results. It may take 10 years or more.
Fuchs, the Memorial Sloan-Kettering cancer research center, led a team to train an AI system that can read "tissue-stain slides." During the training process, researchers need to establish a A huge digital slide library with annotations on the slides, including confirmed and other key data. Each month, the team will make 40,000 slides. Fuchs said: "The number is much more than other people and the task is huge because there are too many variables in the biological field."
Although the Watson project failed, the Anderson Cancer Research Center continues to implement a major project. It wants to collect 1,700 outpatient data, and each patient entering the center will contribute their own data. Andy Futreal, the project leader, said that to make AI systems like Watson work, it is important to combine patient information with research data. He also said: "Once we have obtained the data, we can enter the AI ​​machine learning business to see who is effective and who is not effective under various treatment methods."
IBM continues to get data from partners. In the area of ​​cancer diagnosis and treatment, IBM has teamed up with Memorial Sloan-Kettering Cancer Research Center, Mayo Clinic, Harvard and MIT Affiliated Institutes, and medical testing giant Quest Diagnostics. In collaboration with the Memorial Sloan-Kettering Cancer Research Center, IBM has developed a system that can screen journal articles and form formal treatment decisions. Florida Jupiter Medical Center and a hospital chain in India have introduced systems. . In the discovery of drugs, Watson Health collaborated with the Barrow Neurological Institute, which helped the agency find five genes associated with ALS that were not previously thought to be related to the disease; In collaboration with the Ontario Brain Institute, Watson screened 21 of the most promising drug candidates.
To provide better medical results and reduce costs, does Watson really bring about change? Stephen Kraus, a partner at the Bessemer Venture Partners Fund, believes that there is indeed such a possibility. He focused on the health care industry and invested in many health care-related AI startups. Kraus said: "It's real, not the 'fog' that is used to stimulate stock prices." Like most experts, Kraus is also cautious about unrealistic timetables and commitments, some of which are over-committed From IBM itself. Kraus said: "It's quite difficult. It didn't happen today. It won't happen in another 5 years. This technology won't replace the doctor."
燑br>
1500W Cloth Dryer And Fan Heater
1500W Cloth Dryer And Fan Heater,Ptc Fan Heater,Ptc Tower Heater,Ptc Electric Fan Heater
Foshan Shunde Josintech Electrical Appliance Technology Co.,Ltd , https://www.josintech.com