In 2012, IBM's Watson artificial intelligence software is in the reputation of the most prominent period. It was the previous year, Watson beat two human players to obtain intellectual challenge Jeopardy champions, race, Watson demonstrated superior natural language processing capabilities, amazing. With the popularity of the show, Watson has become the spokesman for artificial intelligence at the time.
In November 2012, IBM Watson recruited as a junior partner, was established in the late 1920s, the Cleveland Clinic will help Watson accepted medical aspects of "training." By training a large number of medical data, Watson can make a lot of "medical assumptions", IBM Watson and health clinics hope to help clinicians and medical students to more accurately diagnose and provide better medical programs. That time, to technology media Singularity "Attendant Doctor Watson" to describe the new Title Watson.
Technically, from the bottom, Watson from the beginning to now are driven by DeepQA. In short, DeepQA is a set of analysis, reasoning and complex software architecture provides the answer, this architecture can read millions of text data, then generate answers through natural language processing technology, to answer the last question in accordance with the scene. It is this architecture to support the development of Watson. 2012, when, Watson can be completed in eight billion calculations per second.
At that time, according to Bloomberg News reported that, IBM had planned to become a commercial area of Watson "super Siri", and the final form is presented with IBM hardware products, in other words, IBM Watson was still want to promote their own hardware business. However, the story up to now, IBM eventually gave up this plan, and to Watson transformed into cognitive business computing platform.
Although IBM did not sell more hardware by means of Watson, but Watson shape or has undergone great changes: from a 10-fridge so much "big man", into four pizza box-sized "Little Dot." Moreover, the user can also tablet PCs, smart phones and Watson connection. Compared to the first generation of products, in 2012 Watson 240% in performance indeed improved, and can handle 28 different types of data, far more than five kinds of previous generations.
2013, Watson's open source API, now integrated into IBM's Bluemix where developers can use this platform, with Watson's "skill" to build a series of artificial intelligence applications.
For the most important step is Watson in 2014, when, IBM set up a $ 1 billion investment in IBM Watson business group, the company has more than 2000 employees, will be fully operational matters relating to research and commercialization of Watson. It has also become a fast-growing start Watson and help IBM cognitive computing into the mainstream media mantra in science and technology.
Time soon to 2016, Watson in many areas have their own base, such as financial advisors, automated customer service representatives, and so on, but we still need to go see first job four years ago, Watson & mdash; & mdash; physician assistants & mdash; & mdash; in the end how to dry?
Now, Watson has been given to the work of the medical aspects of the Health Department Watson. This is another side also show Watson since 2012 the good momentum in the medical field. Since 2012, Watson progress in cooperation with the Cleveland Clinic is not much. Until 2014, the Cleveland Clinic Medical Innovation Summit, IBM announced that oncologists have begun to use Watson to analyze the relationship between the genetic data and medical data to better formation of individualized treatment programs.
Last May, a report based on BI, Watson is likely to have been able to make oncologists "Upload cancer patient's DNA fingerprints into the system, Watson According to this data, the rapid screening of a gene which may lead to mutations, to determine for these possible mutations of the drug. "
In another case, the University of Tokyo researchers used Watson successfully treated a 60-year-old leukemia patient, their approach is the genetic data of patients with tens of thousands of medical literature do comparison, the formation of medical programs customized for the patient. 36kr According to reports, 21 local hospitals plan to use via the Memorial Sloan - Kettering Cancer Center (Memorial Sloan Kettering Cancer Center) Training tumor IBM Watson Solutions (IBM Watson for Oncology), to help doctors provide better patient personalized cancer treatment.
Despite that progress is gratifying, but in the entire medical industry, like this practice is still very small. Last year, iEEE reporter Brandon Keim in an article analyzed several reasons, such as Watson need to adapt to the special hospital work environment, it takes a lot of time to really work on specific tasks and scenarios healthcare environment training Watson.
But this process is very long and arduous training. First, there must be a large number of computer scientists and doctors, pharmacists collect a reference database, and add case study, and then ask thousands of questions. In fact, Watson will not be self-learning, he gives the wrong answer will be researchers evaluated manually adjust Watson algorithm to ensure that the final output of the answer is valuable.
Medical information update speed is very fast, but there may be a conflict between the different pieces of information, it also allows Watson to answer questions before the fast must assess quickly the value of all types of information, different circumstances, to make more accurate judgments.
Watson was unable to meet the people in the medical industry for the slow growth of artificial intelligence to human health huge boost expectations. In fact, almost the medical field of artificial intelligence and the personal computer appear together in the 1970s, but for now, the so-called artificial intelligence-driven personalized treatment, there is still a long way to go.
The current Watson, he must keep up with the current pace of development of artificial intelligence. Since 2012, the depth of learning to become the most popular field of artificial intelligence algorithms, especially deep learning can achieve "self-learning" & mdash; & mdash; unsupervised learning. For example, Google's AlphaGo Go for learning is the depth of learning outcomes to show, through self-learning, AlphaGo grow at an alarming rate, is now the world's top Go "player."
June 2015, "MIT Technology Review" reported, IBM plans to Watson added depth study of technologies, such as translation, voice text of system conversion and more. Today, Watson is still a "hodgepodge" of early natural language processing and analyzing large data sets, and added depth study.
On the other hand, IBM breakthroughs in artificial intelligence, basic research will also have a positive impact on the future of Watson. In early 2016, 8, IBM Zurich Research Center, made the world's first artificial nanoscale phase-change random neurons. This made the 500 neurons array can simulate the human brain works signal processing. This is IBM tremendous progress in the study of "artificial brain" of. IBM said their artificial neural technology and the current development of another artificial neural components & mdash; & mdash; memristor complement each other.
Today, the entire technology industry in transition artificial intelligence, this year due to bot and hottest virtual assistant, whether it is Google, FB, or Microsoft, Apple and Ali, are trying to speed up the layout in the capital and strategic level, Watson seize as IBM stronghold for the future, as well as the transition to cognitive computing core products, the future will continue to play an important role.