* * * Siri is an imagined picture of the future. She is a new way of communicating with computers that will empower the average user immensely. The relationship between humans and machines is constantly evolving, and Siri represents a new paradigm on that trail. Like iTunes and the App Store, she has the potential to change Apple's fortunes, but she also has the potential to become FaceTime - dispensable, or Maps - important but hard to do, or Ping - an underachiever. Acquiring and improving Siri has already cost Apple hundreds of millions of dollars, and in TV commercials, teary-eyed users marvel at her wonders, only to see her become the subject of nightly TV shows that make fun of her. What does Apple hope to get from Siri? Everyone has different expectations. Some think Siri's top five features should include adjusting iPhone settings. Other eager users think Siri should have become the ubiquitous Knowledge Navigator. Of course, a favorite pastime of Siri critics is comparing her to Google Search. siri isn't a sexy librarian. people compare Siri to Google as a matter of course, and it's a lot of fun, but it's not the right comparison. But it's not a good comparison, and Siri will have a hard time beating Google (and, of course, Bing and Facebook) on text-entry searches, since the latter has spent a decade fine-tuning search results with complex and sophisticated algorithms, and has built a huge infrastructure. But Google search has its Achilles' heel: you have to tell Google your intent and instruct it how to build and return search results. This is an opportunity for competitors. Let's say you want to find a vegetarian restaurant in Quincy, Massachusetts, with a budget between $25 and $85, and your Google searching prowess runs deep. Then you would manually type in a very specific string of words, "Vegetarian restaurants in Quincy, Massachusetts, $25 to $85." Even so, you'd still get "About 147,000 results found (in 0.44 seconds)," and you'd still have to filter out the information you want. (The following examples are highly simplified.) This is a "human-guided navigation system". Its navigation environment is a collection of all the data in the world, i.e. the entire Internet. Users basically have to actively communicate their intentions to Google, word by word, and the latter narrows down the billions of "web pages" in this global collection, one by one, and eventually, the user picks out the answer he or she wants from this greatly narrowed down range. Passive IntelligenceBut it's 2012, and our mobile devices are much more 'self-aware' than Google's search engine. For example, a mobile device with GPS, a camera, a microphone, an antenna, a gyroscope, and a variety of other sensors, and dozens of apps with a variety of functions ranging from financial apps to games, already has considerable 'passive intelligence'. It already knows enough about its users to greatly reduce the unknowns of search. ...... If all of this user input and data sensed by the device can be integrated, Siri will have to change the rules of the search game if it is to win the hearts and minds of its users. Instead of a rule-based, linear, largely context-free text search, there will be a highly humanized approach to search: the user will declare his or her intent, but won't need to tell Siri what to do at every step. The user's first step is to start a conversation with Siri, which then puts together an amazing array of services in the background: a precise grasp of the mobile device's location, time, and task. Speech-to-text interconversion, text-to-intent conversion, and handling of the flow of the conversation. Semanticized data, APIs for various services, models for tasks and domains, and, integration of personal data as well as social network data. Let's take a closer look. If you say to Siri, "Remind me to make a reservation for my mom's birthday at such-and-such restaurant when I get to work, and then email me the best route to her house," Siri already integrates with apps and services such as Contacts, Calendar, GPS, geofencing, Maps, Traffic Data, Email, Yelp, and Open Table, which is enough to accomplish this task. Text-only searches like Google's can only do part of the job, and rely on the user to enter a bunch of keywords and make all sorts of adjustments. Now let's try replacing "such-and-such restaurant" with "nice Asian restaurant": remind me to make a reservation for my mom's birthday at a nice Asian restaurant when I arrive at work, and then email me the best route to her house. There's nothing difficult about "Asian" because all restaurant-related internet services more or less categorize restaurants by cuisine. But what about 'better'? What does "good" mean in this context? A traditional search engine like Google would look directly at the restaurant reviews available (that's what they bought Zagat for) for the string "nice", and it might even go a step further and do a combined search like "nice + (romantic or birthday or celebration)" to try to find as many relevant keywords as possible. But we can't fine-tune search keywords across an infinite number of domains, so the above combined searches tend to happen in domains like finance, travel, digital products, cars, and so on. In other words, if you're searching for airfare or hotels, there are a limited number of relevant terms to choose from, and Google has a better idea of what they mean. If you're searching for sheep shearing or 'olive kernel spitting contest', I'm afraid meticulous manual categorization and sorting won't be much help. Context is everythingBut even if a traditional search engine could better select an Asian restaurant for you by associating 'nice' with 'romantic' or 'cozy', you'd still have to think about your wallet. Google doesn't see your bank account records, nor does it know your weekday dining budget and spending habits. So to find really useful restaurant recommendations, it's necessary to set a price range, say from ￥￥￥￥￥ to ￥￥￥￥￥￥, but no more than ￥￥￥￥. This brings us to the boring war between browsers and apps. Traditional search engines like Google have to monitor your clicking habits in great detail to track your transactions if they are to model your buying habits. This kind of monitoring would be so detailed that it would be unacceptable to the user. It's not easy (and probably not legal in many countries), especially if you're not using Google Play or Google Wallet. So Google doesn't see your credit card history and list of bank accounts, but apps like Amex or Chase do. If you allow Siri to communicate with these apps on your iPhone in some encrypted way, you're making a highly selective request, and you're showing that you trust Apple and Siri, so Siri or those apps (and sometimes both together) can be defined as "better" within your budget: no more than $85, 150 to 250, but nothing like that. No more than $85, no more than $150 to $250, and no more than a shabby Chinese restaurant that costs $25 per person because it's Your Tong's birthday. Speaking of Linden, don't forget that under her name in your address book, immediately below "Birthday," there's a customized field called "Food," which says "Asian," "Steak," and "Indian Organic White Tea. On the other hand, you have 37 restaurants bookmarked in Yelp, and without exception they're all vegan-Google doesn't know about that. Your mom doesn't care, but you don't touch meat. At that point, Siri can compare your preferences to give you a mutually acceptable choice. So a simple search goes from "such-and-such a restaurant" to "a good Asian vegetarian restaurant I can afford" because Siri knows - if you ask it - your tastes and your mom's tastes, as well as your financial means: Remind me that it takes milliseconds to look up all of this data in the apps and make up these rules. A lot of your personal information is already cached on Apple's servers. Much of the data in third-party apps is highly structured. Thanks to Apple's agreements with these third-party companies, Siri can use this structured data directly. However, behind Siri's quiet voice, there is still an extremely complex collaborative effort. And Siri's "smart personality" makes ordinary people feel that she is like a magician. Theoretically, Siri's chain of execution can be very long. Look at a common Siri command: If my wife and I don't have anything else on our calendars for tomorrow night, and the event costs less than $50, check the weather for a certain location and the traffic conditions for driving to that event. siri parses it semantically: it translates it into a chain of execution by apps and services: and since she's part of iOS and can communicate with third-party software, Siri has a long chain of execution. And because she's part of iOS and can communicate with third-party software, Siri would be able to do the following hypothetical request: transfer money and buy two tickets, move receipts into Passbook, add a reminder to my calendar, email my wife, update our ****enjoyment calendars, text the babysitter for an appointment, and then remind me about it later.Siri would translate the above into a chain of transactions, which would be executed by Apple's own and third-party apps and services. Siri translates the above sentence into a chain of transactions that Apple's own and third-party apps and services then perform on the verbs and nouns in the chain: Siri syntactically parses a request made up of "natural language" into machine-understandable subject-verb-object constructions, so that not only does it search for documents and facts for you, as Google does, but it can also carry out the request if you authorize it to do so. The request may be explicit, but it may also be implied. The ability to construct deep semantic searches, integrate information from different sources, devices, and third-party software, make rules, and help users execute transactions makes Siri not just a stern-faced female librarian (Google Search), but an indispensable personal butler with certain privileges Siri is the future, and Google understands it, as the traditional search tool has PageRank algorithms and indexes 40 billion documents. After indexing 40 billion web pages, traditional search tools with PageRank algorithms have basically run their course. That's why you see companies like Google buying ITA, the world's largest flight search company, Zagat, cottaging Yelp and Foursquare with Google Places, cottaging Amazon with Google Shopping, cottaging the iTunes Store and App Store with Google Play, cottaging Google Offers, cottaging the iTunes Store and App Store, and cottaging the App Store with Google Offers. Store, Google Offers to Groupon, Google Hotel Finder to Hotels.com ...... and ultimately Google Now to Siri. Google must have the vertical data, knowledge and expertise to better minimize user search intent. Google must have the vertical data, knowledge, and expertise to better minimize ambiguity in user search intent. Keywords, phrases, names, words, derivatives, synonyms, proper nouns, places, concepts, user reviews ...... This data within a vertical can go a long way toward resolving context, scope, and intent. Whether it ends up in search results or Google Now first, Google is indeed desperately trying to build a semantic engine underneath its core services. After all the implicit and explicit third-party services that Google itself and it promotes, 'regular search results' is almost an optional addition. Google has also begun to provide Siri-like 'answers' rather than just endless 'links'. If you search for "Yankees" while watching MLB baseball, you'll get real-time scores first, not the history of the Yankees or their newly constructed stadium.Siri is the "high maintenance woman. They hired a bunch of PhD students to design the algorithms, bought tons of servers, built data centers, and optimized and optimized. In terms of advertising revenue, the input-output ratio of search is unprecedented in the history of the Internet. Apple's investment in Siri is much shorter and less visible to everyone. It would be suicidal for Apple to take on Google in the "find what you want" category, but is it possible for Siri to grow sustainably and eventually mature? Apple's programs are basically the only ones that can support themselves. Advertising isn't Apple's strong suit, so what are the other possible business models for Siri? In 2014, Siri will be available to about 500 million Apple users, and if Apple can get half of them to use Siri to generate 12 transactions per month (e.g., $1 per transaction, with 30% going to Apple), that's a billion-dollar business. Optimistically, that's a lot more than $1 per transaction, a lot more than 12 transactions per user per month, and a lot more than half of all Siri users making transactions - especially if Siri is open to third-party developers. These assumptions are, of course, a figment of my imagination, but even at the most conservative estimate, the revenue generated by transactions is significant. Don't forget that digital music and movies alone make iTunes an $8 billion business, and Siri, which was a simple text-based search service before Apple acquired it, has evolved into a search service that uses voice recognition/dictation as an input interface. She's now evolved into a search service that uses voice recognition/dictation as an input interface, and then she'll go on to become a "conversational interface" for transactional tasks, which will be much more interesting and accessible to hundreds of millions of ordinary, non-computer savvy users. Siri as a Transaction Machine Transaction-centered Siri has the potential to shake up the $500 billion global advertising industry. Pure" information is preferable to fleeting advertisements or a bunch of search results that need to be scrutinized by the willing consumer, and Siri, rooted in a personal mobile device with a high degree of contextual understanding, can deliver unparalleled relevance of "pure information" when the user needs it most. Siri cuts out all the middlemen and connects the customer directly to the merchant without Apple itself having to be involved in the transaction; all Siri does is match the user's intent and provide choices, and it does so more accurately, proactively, and reliably than any large-scale product of its kind we've ever seen. Siri also has another advantage over image or text ads: the transaction can be for something other than money. It can be discount coupons, Passbook vouchers, frequent flyer miles, virtual goods, game score rankings, check-in points, credit card credits, iTunes gifts, school course credits, and more. In addition, Siri doesn't require an interactive screen to communicate with users and accomplish tasks. Apple is using Eyes Free technology to bring Siri to voice-activated systems. First for in-car systems, and then perhaps for other embedded environments that don't require a visual interface. Since Apple has the largest and most profitable app and content ecosystem in the world, and 500 million credit card users, the value that Siri's "transaction" concept can provide to both users and merchants is greatly enhanced. Too early? Too late? Or simply immature? Despite the promise, there are a number of unknowns about Siri's future. Here are a few of the major hurdles: Performance: Siri works in the cloud, so any network latency or disruption will take away its power. This is a difficult hurdle to overcome, as vertical knowledge must be aggregated from thousands of users and orchestrated in the cloud through partners' servers. Context: Siri doesn't just have to understand grammar, it has to work horizontally across many verticals to understand context. Apple currently sells iOS devices in more than 100 countries, and Siri must eventually be able to understand the languages of all of these countries and deal with the complexities of cultural differences, as well as connect with local data and service providers. Partners: Choosing data providers, especially overseas, and maintaining service quality is not simple. In addition, Apple will face bids from Google and other rivals when buying data. Scope: As Siri rises in stature, so do user expectations for its accuracy. Apple has been carefully and slowly adding to Siri's knowledge in various popular fields, but the "Why can't Siri answer my question about [insert a cold field here]?" type of complaint is bound to come up. Operations: As Siri grows, Apple will have to greatly expand its workforce. Not just engineers specializing in semantic search and AI, but people in data acquisition, entry and error correction, business development, and sales. Management: After Siri was acquired by Apple, two of the three founders left and Tom Gruber stayed on. Apple recently tapped William Stasior from Amazon, who was previously the CEO of the A9 search engine, to run Siri, but collaboration with data providers is as important to Siri as engineering capabilities, and A9 is an old-school search engine, not the same as a semantic platform such as Siri.API: Third-party developers are clearly looking forward to the day when Apple opens up its Siri API. API: Third-party developers are clearly looking forward to the day when Apple opens up the Siri API. For Apple, opening up the Siri API is both a gold mine and a minefield. Because the same or similar data can come from multiple third-party data providers, opening it up to third-party software developers could easily create operational, technical, and even legal headaches. Regulation: If Siri is a success, competitors will likely call on the U.S. Department of Justice, the Federal Trade Commission, the Federal Communications Commission, or their European counterparts to intervene and slow down Apple's progress by claiming that bundled software creates unfair competition until they can catch up. Obviously, an avant-garde platform like Siri can't be without its share of problems and risks. Apple's last two successful Internet products - the iTunes Store and the App Store - used technology from a previous era and included operational problems. Newer Internet products such as MobileMe, Ping, Game Center, iCloud, iTunes Match, and Passbook aren't big hits. Still, Siri is a landmark opportunity. For Apple, she could become a transaction-based cash cow; for users, she represents a new way to search for information and complete tasks that is more approachable than anything else so far, and Siri's success or failure depends on Apple. (Translated by Li Ruyi)-1. The original text is high-maintenance lady. high-maintenance means a person who is demanding, emotional, and in need of constant care from others. "High-maintenance lady" is Yishu's private translation, which pays homage to Yishu.