From 16 years on 18 AI to replace the manual flocked to the topic, but to tell the truth, to evaluate the current artificial intelligence, or AI, then most excellent, I can only say that they are great imitators, rather than thinking By. Whether it is from now on AlphaGo to AlphaStar by imitating humans are operating force reached at the above calculation, and why they are completely unaware of this operation.
If you need to explain the operation of AI logic at work, that machine translation is a very typical example, in fact, long before the emergence of the computer, Soviet scientists have put forward the idea of the foundation, after the ENIAC, IBM, led by companies are also involved into the machine translation studies. Many scholars are looking forward to, like a science fiction film in real-time translator that can occur as a tool, then the world will not appear speech disorder, it is bound to be more united.
Translation of this matter in the eyes of many seems not a difficult thing, eleven control word word translation is not on the line, in most languages, has a corresponding vocabulary, "Mom" corresponds to "mother", "Tables "corresponding to the" desk ", you can establish a corresponding relationship; linguistics has also been a mature system, the grammar rules of the language features do dismantling, become effective algorithm strategy, would not it should be a natural.
But it did not, a lot of slang words and habits are in actual human use of language occurred during no law, the interpretation of the rules by the way, is almost impossible.
For example: Less bread No jam (London Transport )..
This is the London station's slogan
literally:
less bread, do not jam.
In fact this ad clever use of rhetoric semantic pun, because in informal English, the word bread have "money" meant, but jam the term is a "traffic jam" means.
So, I want to express the meaning of the ad is:
save money while riding the London Underground, provincial, rain or shine.
Although it realized the significance of this translation between the original and the translation peers, but its effect is not satisfactory in that Pun lost, indeed fly in the ointment, is "obviously know the word, but that translation does not come out."
A language and another language is not a simple word corresponding relations, cultural system is the obstacle, so this event towards the weak AI and strong AI, alone translation, the machine far from able to replace humans. Despite language translation, then the AI will make what work has no value. Simple predictions, the so-called data have improved, with clear objectives, based on existing experience and work carried out, it will become worthless.
Such as market trafficker, the main content of the work is to try to put all the available channels, and calculate ROI, and continue to enhance investment in high ROI channels, reduce investment in low ROI channels. Secondly, operational activities staff responsible for configuring various activities. Several activities are already forming a template, he will fluctuate based on each user's data, from their own experience to judge what kind of activities should be selected, as well as how to configure specific rules at the event.
These two is typical, it will quickly be replaced by work. If you are doing this kind of work, but also look alert. And like product decision-making and operations often do not have the conditions "large number of iterations and iteration result data", in particular a decision may be followed by a lot of constraints, in a coupling factors.
In such a scenario, AI may substitute product manager: do short video can be automatically combined the thousands or even tens of thousands of products and solutions available to users, based on the user's habits, to find the best solution, continue combination of automatic continuous iteration.
Before having such flexible enough production and testing conditions, product manager also unlikely to be replaced. But now look at the development of Internet Product Design patterns have been moving towards such a future.
Internet companies are in line with more than two use cases of the A / B test, with a comprehensive statistical analysis of data buried point to find a solution, though not intelligent, but are gradually replace the work of product managers, product managers make it becomes increasingly important.
Product manager like rule-based products do, and now the gradual emergence of a lot of decisions are based on statistics. Do you feel like a history with machine translation?
And what kind of work to make the AI more valuable?
In "AI minimalist economics" in the view that data, judgment and action are complementary products, it becomes more important. What are these three is it?
Data, statistics do not refer to the person or people who do data analysis, but the person who owns the data / organization and have the ability to get data of person / organization. For example, based on the recommended consumption statistics become increasingly important, and that Ali, Jingdong products and data accumulated behind it becomes important; become increasingly important statistics-based unmanned, that the transport sector, navigation software and networks about car platform becomes important.
There are a lot of people who work for data, but is in fact doing the work AI alternatives, such as simple navigation function, the role will travel data based on past user to do the analysis and statistics in order to find some rules. But powerful enough AI do not need rules, as long as the input is sufficient, you can give much better predictions, it will put such a role replaced. (Actual scene may be much more complex, the need to make a judgment, still requires human intervention.)
Judge, referring to the AI to tell what is right and what is wrong.
AI learning and mastering a capability for the service of humanity, and that the "purposes" and that people must be given. Here it is necessary to give people judge, like telemarketing sales industry inside robots, robot phone really easy to use is not reflected in simple machines to read and implement feedback for big data, because simply rely on to achieve human-robot interaction impossible, so need to assist customer service personnel to complete, from the point of view seems to not replace this direction, become more assistance, but also significantly improve the efficiency of all whether it is above or quality of the above.
For example, the navigation function just mentioned, you can use machine learning to do is smart enough to be counted as AI products. But this does not require any representatives to participate in product operation yet? Of course not, this product for the target to be achieved in particular constraints, is still to be given to the people.
Of course, the shortest time is the first element, but the extent to which users are willing to pay in order to save time and high-speed charge? The extent to which users will hate traffic jams or traffic lights? And so on, these people feel based on the judgment, can affect the recommended route navigation strategy.
Action, of course, is based on the predicted executed. Get a good algorithm unmanned, how to apply? Get the recommended consumption behavior, how to apply? However, the content itself is usually performed will extend out whether there is likely to be replaced by AI problems.
Expect conditions to gradually mature, there may be a superior product decisions based on statistical AI future, just like in 2005 Google rounded out the field of machine translation, as still quite a sense of crisis.