In the early seventeenth Century, Japanese businessmen, active in South China, Southeast Asia, and the Java islands, believed that their warrior knives were about to become the most powerful arms trade in the world.
At that time, the central and southern peninsula nobles and city states in the scuffle took out real gold and silver to buy Japanese samurai knives coming from far away. In Siam, a Japanese Knights knives will be sold to 42 silver pieces, five times the length of the local knives. With the strong casting technology and the oceangoing fleet, Japanese businessmen seem to see a new era.
A short time later, the Spanish fire rope gun appeared on the Asian market, also selling four two silver one
This sad story tells us that when you think you are at the top of the era, maybe the times are ready for you.
Of course, this article is not to discuss the samurai sword, but to talk about a very famous AI project: IBM Watson.
In 2011, IBM's cognitive computing system, Watson, was born, beating humans for the first time on a Q & A show. Then the value of the rising Watson gradually became IBM, and even the global AI project representative. There was no DeepMind, no smart speakers, no in-depth learning framework.
Then, all the things on the upper side have.
About Watson's latest news, it was not long ago that the media broke the news that Watson health department had to fire about 50% to 70% of its employees. And it's obvious that the gang is already on the ground, because it is based on a continuous decline in income for the IBM five years, a continuous downsizing, and less than 0.1% of the return on shareholders.
How on earth did Watson, the darling of the halo, come to
Before attempting to answer this question, we must correct this understanding. Many analysts believe that the IBM Watson's embattling is a very representative problem in the AI industry: it shows the common ground problems facing AI applications.
Of course, the AI industry is facing the problem of landing, but this judgement is too one-sided, just like it can not be said that the defeat of the samurai sword is the failure of the weaponry history. The rapid rise of Watson and its long decline are all due to its uniqueness: premature advance and then premature lag.
An untimely lead elder brother
Back in 2011, Watson, who won the championship in the answer shows, was very proud. With the screaming of the newspapers and the screaming of the capital market, the public seems to believe that AI in the movie is coming to the world.
The effect is not only because of an answer competition, but also by the decades of IBM in the new world of technology, as well as the belief in the product form of the quiz system in the European and American markets.
Since 80s, the expert system has become the main product form of AI Renaissance. At that time, the AI community generally believed that the future AI was to build an omnipotent expert, and then humans could always ask for advice. This dream lasted for several decades and was ultimately proved by IBM to have a AI system that could be imagined beyond the best results of a human high hand in a question and answer contest.
The story is like the predecessor of AlphaGo. But the difference is that AlphaGo is just a technology demonstration. DeepMind and Google do not plan to make money directly.
But IBM at the time seemed to believe that the universal solution system really could be commercialized, and that it had made massive technology and team acquisitions. Finally, in 2014, IBM formally created the Watson business department and started a commercial journey.
Many people know that IBM scientists have named the technical system behind Watson, called cognitive computing. They believe that human beings can calculate cognition by means of NLP, emotion recognition, machine learning and so on. That is to say, we can find the rule and truth behind the complex unstructured data. This scientific mode can be released to various industries and discover the golden rule behind all kinds of commercial and technological behaviors.
This logic sounds very much the same, and it has proved its effectiveness in some fields.
So we saw a variety of institutions and businesses buying Watson services over the years, especially after the Watson health program started, and many hospitals and medical institutions wanted to use this new technology to explore the secrets behind the clinical medicine.
No doubt, before the five families of Google, Amazon, Facebook, Microsoft, and apple are not fully AI, Watson is the most popular AI project in the world, and all the time is the representative of the AI category of cognitive computing.
But with the rise of machine learning and interactive AI, rising stars have climbed another technology tree. Everyone seems to have ignored the cognitive computation.
It was a little embarrassing. The predecessor was out of a road. When it was in a good mood, it was found that the younger brothers, who were supposed to keep up in the back, had gone from the other road. The unique leader has become a unique conservative school, which is similar to the story of samurai sword.
After entering the eighteenth Century, the large-scale Knights trade was no longer competitive.
The knife is very sharp, but the battlefield doesn't need it
Is the AI ability of IBM not strong? On the contrary, IBM has always shown a strong absorptive capacity for AI technology, whether it is in the underlying R & D or project acquisition. Only in the past two years, with the large-scale rise of AI, IBM began to lose its place in the competition between talents and new markets.
Compared to the technical problems, what really causes embarrassment is the Watson's assumption of product logic and its poor market reputation.
Of course, commercial service AI technology has so far not decided the real outcome. Google and Amazon's business AI services are also often spitting. The problem with Watson, however, is that it is hard to use as the opposite of other platforms.
Whether it is the business service system or the health market, we can often see Watson encounters such slots:
1, aim high
In 2015, when Watson was killed in health care, it was said that Watson would benefit one billion people to solve, diagnose and treat 80% of the 80% types of cancer.
There is no doubt that 3 years have passed and this promise has not been fulfilled. This is a routine routine of Watson. First pick up the most difficult problem and publicize it. When most of the AI applications are focused on medical image interpretation, case history management and so on, Watson has been heavily involved in cancer research.
The basic scheme of Watson in cancer research is to rely on the input of a large number of real medical records to interpret the causes behind the cancer, the deep pathology, and give advice on the treatment. But it should be noted that almost all medical institutions that have purchased Watson have clearly pointed out that the role of this system is to help research.
In fact, the clinical decision-making of cancer treatment is very complex, the advanced degree of machine learning is difficult to support the solution of the core problem, even the data chain is difficult to build clear. With more and more medical institutions announcing their withdrawal from cooperation with IBM, Watson's medical trip seems to have been questioned.
On the whole, AI medical treatment is more suitable for trivial and repetitive medical work today, such as copying a medical record, looking at a X ray, checking a laboratory test, and so on. These work can liberate the time of the doctor. Let AI itself treat major diseases, both technically and safely.
The tendency to aim too high is not just a health issue for Watson. Watson is more like a robot strategy consultant for the boss in the area of enterprise services, while AWS replaces and helps to perform a specific jo
2, harsh application conditions
Since Watson is positioned to solve major problems, it means that it requires complex training and data entry process.
If we just want AI to replace the face in the video, then it's just input video and replace photos to train. And if you want to use Watson to provide a major business decision proposal, and even reveal the truth of a certain kind of cancer, the data needed can be extremely large and complex, and even a large number of experts need to be carefully studied and screened.
With this product training feature, it is destined that Watson cannot be
In February 2017, the MD cancer center of Anderson, a famous cancer research institution, suspended cooperation with IBM Watson, one of the reasons is that the cost is too high.
3, a firm closed policy
Ai is becoming more and more open source, more and more popular.
Since its inception, Watson is a closed system. Companies only get beautiful UI interfaces and final reports, and other researchers can't get anything. This is obviously something to share with today, and the more the AI community shares, the more it will not fit in.
Especially in the industry research platform, open source framework and Documentation means that a large number of practitioners can share the development results on the platform and make the whole system alive. Today's platforms such as Google and Facebook are sticking to this logic, but IBM Watson has never been interested in open source.
Today, when a AI project locks up the gate, it seems to lock itself in the meaning of the previous era.
No matter how fast a samurai sword can be adapted to a battlefield dominated by gunpowder,
Cognitive calculations that fall into cognition
There's a door called
Why do Watson's negative news often experience the mockery of the tsunami? Especially in the American media?
Of course, there's a gun for the bird, but the bigger reason is because of the aura that Watson wore to it.
Interested friends may wish to search and search. In the past few years, Watson's advertisements in Europe and America are enough to become a AI movie to save the world. Add IBM's fearful marketing formula, and exaggerated speech and film style. From 2014 to 2016, the purchase of Watson became synonymous with cutting-edge and fashionable.
As for the use of things that are delivered to the company's hands, that is what they say.
This kind of marketing strategy, rather than the promotion tactics that belittled the application effect, once caused the public to have such an illusion: Watson is a universal magic box. Enterprises can let it solve all problems, and the application of Watson will directly start a big era.
If it is today's Chinese consumers, it is obvious that this PPT mode has long been immune, but a few years ago, Americans obviously still need to be honest. No matter whether IBM really believes them, they at least succeeded in letting media and customers believe in the myth of Watson. Adding to the medical field to solve the problem of cancer, a short - term stimulant enough, Watson is under the technical surface of the European and American markets, floating too much of the non AI business impulse.
And when the myth begins to break, the laughter will certainly be louder.
In fact, doubts about Watson's real enterprise service capability always exist. Although Watson seems to be able to provide a lot of service and decision help for the enterprise under the fancy interface, if it is broken down, many of them can use free data services to get it, and decision making advice is more intelligent.
The final customer may be surprised to find that the original cognitive computation is not calculating the cognition of AI, but calculating my cognition. It's a little embarrassed.
In a word, hard core AI technology and small public product ideas, plus excessive marketing and unsatisfactory application effect, IBM Watson into a cup of unique cocktail, a very beautiful but lost the merchant's knives.
With the shadow of the AI expert system in the 1980s, it used the proud IBM promotion of the 1990s to live in the era of machine learning with the themes of open source and equal rights.
After all, the age is too hard, and sharp IBM can not stop the time that rolls away.