Dawn is checked compile from IEEE Spectrum
Quanta report | Public date QbitAI
Begin from 2011, blue giant IBM went up to be noted again in the detain on AI medical treatment, established a luxuriant headquarters for its AI branch even. All these is to his lead fixed position holds in AI tide.
However the medical treatment technology of IBM Watson had used 8 years, never leave deep impression to the person however.
Eliza Strick of IEEE Spectrum advanced subeditor thinks much cry and little wool of IBM Watson medical treatment, early days gave very tall consent, there is not telesis far however nowadays.
In the tentative idea of IBM, the memorial library of Watson is mastering the knowledge of all sorts of rare diseases, can settle a stubborn case inside a few seconds. If Watson can head this kind of instant and professional knowledge the hospital of world each district and clinic, artificial intelligence can reduce diagnostic mistake it seems that.
But the achievement of Watson domain of research of be confined to. One of the biggest challenges that IBM tries to apply Watson at medical bound when cancer, encountered machine study between means and doctor work way match far from.
8 years acceptance fails to cash
2011, IBM Watson is in intellective contest TV program " critical margin " in conquer two mankinds champion, subsequently IBM announced his ambitious plan: Future lets Watson become Doctor AI. IBM returns acceptance to be in future rolls out product of first medical treatment health in the 18 time to 24 months.
Actually, the project that IBM announces at that time did not produce any commercial products.
After this in 8 years, IBM high-key him advocate is in the effort on technology of AI medical treatment, but the great majority of project of AI medical treatment that cooperates with other hospital, college has failed.
According to IEEE Spectrum statistic, begin from 2011, 25 of IBM Watson and other orgnaization collaboration in having representative project, only 5 collaboration project rolled out product of AI medical treatment.
IBM still tries to strengthen oneself through buying, but in-house personage expresses, bought company did not have too big effect.
The product that comes out from the Watson Health branch of IBM and Doctor AI that once imagined differ completely: It more resembling is the AI assistant that can carry out certain and day-to-day job.
California university san Francisco divides school medicine to fasten director Robert Wachter to say: IBM encountered a trouble. They put sale in the first, the product is put in the 2nd, lift up or let down with a rope became enough the appetite of people, but after project mount a horse, discover this is problem of a special scabrous again.
NLP is short-term cannot solve a problem
Current, the United States superintended an orgnaization to approve tool of a few AI to be used at the hospital only. These products that start a gender basically are in visual domain, and IBM devotes oneself to to solve medical treatment problem with NLP all the time however.
Company of other AI medical treatment is considering to analyse image with computer vision, if X mating plate and retina scan. But the product with video without such analytic medicine IBM, although they have an active research project in this domain.
Eliminate medical treatment is video, even if is today best AI also understands complex medical treatment message hard. With software professional to human doctor knowledge undertakes encode is a very difficult issue. What whole world of no less than sees, IBM had acquired these painful lessons in the market.
Although IBM did not abandon a goal, but its failure has made clear to technologist and doctors, founding Doctor AI is how difficult.
Watson was in 2011 the victory on intellectual interlocution, proved its outstanding performance in NLP respect. To play this game, it must analytic and complex paronomasia clew, search many text database, find and determine best result.
"Look Watson is almost understandable the meaning of the language, is the pattern that identifies a word not just. " ever held the position of IBM academy 2011 Martin Kohn of presiding medicine scientist says.
Be in early actually attend " critical margin " before, IBM has considered AI iatrical possibility.
Much patient data suits to be used at AI medical treatment to consider very much it seems that, when beginning to use electronic health to record with the doctor when the hospital especially.
Although data of wh some of which can be digested easily by the machine, exemple if lab result and life style are asked for,measure data, but still having much is " blame structuralization " information, the doctor takes notes and leave hospital for example record. This kind narrates sexual text to occupy the content that records about 80 ％ to the patient.
Kohn thinks, the NLP ability with powerful Watson can transform for medical theory. Watson can read the entire content of healthy record of the patient and medical literature, include schoolbook, person of the same trade to evaluate detailed list of medicaments of periodical article, approval to wait. Pass a visit all these data, watson may become a super doctor, can discern the pattern that gives the mankind to cannot see.
Department of Colombian university medicine teachs Herbert Chase to ever spread out collaboration with IBM, hope Watson can become a kind " clinical diagnostic support " tool, they studied the prototype of a kind of diagnostic tool together. But both sides parts company from 2014, the reason is Professor Chase feel slow to Watson progress delay disappointed.
Chase teachs one of numerous co-worker that are IBM Watson only, professor of radiology of university of another co-worker Maryland thinks, AI tool is opposite inside 10 years is indispensable for the doctor, but he does not believe IBM can come true they.
He says: "I do not think IBM is in the foremost edge of AI technology, the most exciting thing happens in apple, Gu Ge, Yamaxun these companies. The most exciting thing happens in apple, Gu Ge, Yamaxun these companies..
2014, kohn left IBM, he says: "Proving you have powerful technology merely is insufficient, still need to prove to others you can make useful thing. Still need to prove to others you can make useful thing..
The remedial effect that the paper that Kohn is awaiting medical periodical to go up all the time proves to AI can improve a patient saves the expenditure of medical treatment system. So far, what this kind of article publishs is very few, watson also does not have any results.
IBM first domain of sortie AI medical treatment
When trying to introduce AI medical treatment, IBM is facing enormous challenge, it has lagged behind at the tycoon such as Gu Ge, apple in a lot of computation domains. IBM needs a few important thing ability to assure a position. For this, IBM invests a large number of capital in domain of AI medical treatment.
2014, IBM invests 1 billion dollar to Watson branch. 2015, IBM announces to establish a special Watson Health section, to 2016 metaphase, watson Health had bought company of 4 healthy data, total expenditure restrains 4 billion dollar.
IBM had the technology that lets AI enter place of field of medical treatment health to need, resource and commitment it seems that. "Using AI at medical treatment health is a formidable task, this is a challenge, but we are being done so. " IBM cognitive solution and IBM academy are advanced vice-president John E. Kelly says. Begin from 2011, kelly is in charge of coaching Watson studies, what he still is in charge of pair of Watson Health now is direct supervise.
He says the company has gotten lost when need: "We are continueing to learn, so our product can learn as ours and change. So our product can learn as ours and change..
The medical treatment of IBM diagnoses a tool is an example, it did not appear on the market, IBM Medical Protection and life science study vice-president Ajay Royyuru says: "Diagnosing is not the field that we should enter, this is the place that medical treatment experts can become very well, no matter you use AI to be done much weller, won't replace human expert. Won't replace human expert..
To find case of AI iatrical trade, IBM developed a series of items that make a person dazzling, participator differs in be aimed at Medical Protection system: Doctor, administrator, insurance company and patient.
In making a variety of attempts, the NLP that IBM tries hard to let Watson goes understanding medical treatment text. But Yoshua Bengio of professor of Montreal university computer says, in medical text file, artificial intelligence system papers a gender without legal principle Jie Jimo, the delicate clew that doctor of unintelligible also mankind notices. So far, do not have artificial intelligence to be able to achieve the understanding of human doctor and insight.
IBM is one of main challenges that this company encounters to the research of cancer. Mark Kris of a lung cancer expert expresses: "I think to be able to know this without the person (implementation cancer is diagnosed) can beautiful how long. " the research organization that he is in was mixed early 2012 IBM Watson spread out collaboration.
Kris and other doctor trained system of an artificial intelligence 2015, this system made Watson tumour product. The doctor of center of cancer of university MD Anderson and IBM collaboration founded Dekesasi the name is Watson tumour adviser. Leukaemia branch checked this tool, but it never makes commercial product.
Watson tumour adviser was sufferred criticize with all possible means. The article points out Watson tumour is the proposal that no use can offer risk even. Kris says he often hears criticism to say this product is not " true artificial intelligence " . IBM and the cooperative project that MD brings center of heart dark cancer cost 62 million dollar, end in order to fail finally and be cancelled.
This project announced the reality of the acceptance of machine study and Medical Protection matchs far from, in " true AI " be opposite with current doctor existing between the requirement of functional sex product difference.
Watson tumour is recorded through getting health of much cancer patient and medical literature learns. IBM hopes Watson depends on its powerful computation ability, the examination these record hundreds medium variable, discover the pattern that the mankind cannot see. To oncology home, this sounds nurse like cancer the potential breakthrough of the respect. To IBM, the product that it sounds great like.
The implied meaning that Watson learned very quickly how to browse clinical research determines basic outcome. But it won't read an article in that way like the doctor. The doctor is collect information from inside the article, treat a method with what will change them.
And the thought of Watson is to be based on statistical data, collect the statistical data that concerns main result what so it can do namely. "But the doctor is not done so. "But the doctor is not done so..
For example, 2018, FDA approved medicaments of a kind of new cancer, it is specific to all showing the tumour of gene mutation is effective. Human doctor can suggest patient of every lung cancer accepts this kind of gene to detect, but Watson won't change conclusion according to having patient of 4 lung cancer in 55 patients.
Not agree with with level of system of existing medical treatment
Watson cannot extract his opinion from the new anamnesis episode in medical treatment document, this is the first pace that it fails only. Researcher returns discovery, it also cannot resemble anticipating dig information from inside electronic health record of the patient in that way.
Researcher lets Watson study the health of leukaemia patient is recorded. In the paper that was published at The Oncologist 2018, this group report says, adviser is in the oncology expert that is based on Watson to recorded field of information of the extraction in medium text documentation to gain the success of different level from medical treatment. Diagnose in processing when waiting for clear idea, accuracy differs to 96 ％ from 90 ％ , but the information to relying on time (if treat time line) , accuracy comes for 63 ％ only 65 ％ .
Not only such, IBM realized the dream of Doctor AI super to still encounter another retype attacks: Researcher realizes, watson cannot will new patient and before the cancer patient that discovery conceals mode undertakes comparative. The expert oncology that IBM hopes AI can imitate them originally domestic ability, the result before using and experience, formulate remedial strategy for new patient.
But the study that the level with Medical Protection current system does not encourage this kind of real world. The proposal of AI system is based on the mode that it discovers in medical treatment record, but if do not have the strict control that studies scientificly, this kind of discovery can be considered as dependency only, is not causal.
IBM thinks the standard of Medical Protection must be changed, ability lets AI come true its are all latent capacity and translate into medicine.
Tumor expert distrust
Reportedly report, IBM searchs product of its Watson tumour in the United States the trouble was encountered when buying the home. A few judgement that oncology home says to they believe themselves, do not need Watson to tell them what to should do. The level that someone says to Watson just suggests they are very clear treats a method.
In a few years of in the past, the hospital that cooperates with IBM already began to publish about Watson tumour the research of relevant experience.
In India, the doctor of center of Manipal integrated cancer is ill to 638 breast cancer in the exemple, watson and the consistent rate that expert cure suggests are 73 ％ . Watson is behaved in center of medical treatment of Gil of Korea Gachon university poorer, this hospital is the optimal proposal that patient of 656 colonic cancer offers, only 49 ％ and expert photograph match.
Watson Health also has a few successful case. IBM and the partner that north blocks Luo Laina university published the first paper about effectiveness of Watson For Genomics 2017 in, attend the cancer patient of the 32 ％ of this research, watson discovered the potential and important mutation that the mankind never discovers.
The story of Watson Health is a caution, everybody likes lofty aspirations and great ideals, but nobody hope is put dove.
, be over,
Cordial invite applications for a job
Quanta head of QbitAI · name label makes an appointment with an author
? ' ? ' ? Track new trends of AI technology and product
