Chapter 12 6. Less is more experienced British butler

In December 1980, Weissner and I were guests at the lovely country house of Nobutaka Shikanai (Sankei Shimbun, former chairman of Fuji TV).The villa is located in the Hakone area of ​​Japan, not far from Mount Fuji.We are convinced that Mr. Shikanuchi's newspaper and TV media empire will benefit greatly from participating in the establishment of the Media Lab, so he will be happy to support the establishment of the Media Lab.We further believe that Mr. Kanuchi's personal interest in modern art will coincide with our dream of trying to integrate technology and artistic expression, combining new inventions with creative applications of new media.

Before dinner, we took a walk to admire Mr. Kanuchi's famous outdoor art collection, which during the day is the Hakone Open Air Museum.When we had dinner with Mr. and Mrs. Shikanai, Mr. Shikanai’s private male secretary was also with us.Mr. Shikanai knows nothing about English, but his secretary can speak beautiful English and plays an important role in our communication.Weissner first opened the conversation, saying that he was very interested in Calder's works, and then introduced the Massachusetts Institute of Technology and his own relationship with this great artist.After the secretary listened to the whole story, he translated it in Japanese from beginning to end. Mr. Shikanuchi listened carefully.Finally, Mr. Kanuchi pondered for a moment, then raised his head, looked at us, and let out a sound like a shogun.

The secretary then translated: "Mr. Shikanai said that he also admires Calder's works. The works he bought recently are..." Wait a minute, did Mr. Shikanai say these words? This happened again and again throughout the dinner.Weissner said a few words first, and the secretary translated it all into Japanese. Mr. Shikanuchi's answers were almost the same "Oh-oh--", but the secretary was able to translate a lot of explanations.That night, I told myself that if I were to build a personal computer, it would have to be as capable as Mr. Kanuchi's secretary.It has to be able to understand me and my surroundings in such a fine-grained way that it can automatically expand or compress the signal, so that most of the time I'm superfluous.

The best metaphor I can think of for a human-machine interface is that of the seasoned British butler.This "agent" can answer the phone, identify the caller, interrupt you when appropriate, and even make up white lies for you.This agent has a knack for timing, getting the timing right and respecting your idiosyncrasies. Someone who knows the butler has a lot more to offer than a complete stranger.That's really nice.Einstein can't help but very few people can enjoy the service of this kind of human agency.A similar role that we see more often on weekdays is the secretary of the office.If the secretary knows you and your work well, he can represent you very effectively.

If one of the chief secretaries fell ill, even if the temporary employment agency sent Einstein, it would not help.Because the important thing is not IQ, but whether there is consensus between each other, and whether the secretary can use this consensus in your best interest. Until recently, making computers capable of such capabilities was a distant dream, so many people didn't take the concept seriously.However, the situation is changing rapidly.Now, many people think that such "interface agents" is a feasible idea.As a result, the idea of ​​an "intelligent agent" that was rarely asked in the past has now become the most fashionable research topic in the field of computer interface design.It's clear that people want to delegate more functions to their computers than to do everything themselves.

Our idea is to design a knowledgeable interface agent that understands not only things (the flow of something, a certain area of ​​interest, a certain way of doing things), but also your relationship to things (your tastes, tendencies , and who you know).That said, the computer should do double duty, just as cooks, gardeners, and drivers apply their skills to your tastes and needs in food, gardening, and driving.When you delegate those tasks to someone else, it doesn't mean you don't like cooking, gardening, or driving, it means you can choose to do those things when you want to, and because you want to, not because have to do.

The same is true of our relationship with computers.I'm really not interested in entering a system first after surfing the Internet, and then going through a bunch of communication protocols to find your Internet address (address).I just want to pass on the message to you. Likewise, I don't want to be forced to read thousands of bulletin boards just to make sure I haven't missed anything important. I want to let my interface proxy do it for me. There will be many digital stewards, some living online, some right next to you, in the central systems and peripherals of organizations (whether large or small).

I have told people that I have two lovely smart pagers.It was brilliant at delivering important messages to me at the right time in flawless English sentences.My approach is to have only one person with the pager number through whom all messages go, and only he knows where I am, what is important, and who I know (and their surrogates).The intelligence comes from the headend of the system, not the periphery, nor is it on the paging body. However, there should also be intelligence on the receiving end.Recently, the CEO of a large company and his assistant visited. This assistant carries the boss's pager, and he will remind the boss of some urgent matters at the most appropriate time.The assistant's sophistication, timing and prudence, will be engineered into the pager's function in the future. "My Daily"

Imagine if an electronic newspaper could be delivered to your home in bits, and assuming those bits were transmitted to a magical, paper-thin, flexible, waterproof, wireless, light and bright display, it would take If you want to design an interface for this newspaper, you may need to use years of human experience in making headlines, designing layouts, breakthroughs in printing, experience in image processing, and many other technologies to help readers browse and read.Done well, it can be a great news medium; done badly, it can be horrible. We can look at a newspaper from another angle, which is to regard it as a news interface.Digital life will change the economics of news selection. You no longer have to read what other people think is news and what others think is worthy of a page. Your interests will play a more important role.Articles that were abandoned in the past due to concerns about the needs of the public and could not be arranged on the layout can now be used by you.

Imagine that the interface agent of the future could read every newspaper, every news agency on the planet, master the content of every broadcast TV, and then assemble the data into a personalized summary.Only one unique edition of this newspaper is produced each day. In fact, the way we read the newspaper on a Monday morning is very different from the way we read it on a Sunday afternoon.Scanning the newspaper at 7 a.m. during the workday is just a matter of sifting through information, extracting the personal bits from the common bits sent to thousands of people. Most people will throw a full-page newspaper into the trash without even looking at it, and browse the remaining pages for a while, and there are very few parts that are really carefully read.

What if a newspaper company was willing to let all the editors edit a newspaper at your beck and call?This newspaper will combine the headlines with some "not so important" news, which may be related to people you know or people you will meet tomorrow, or about places you are going to and just left, and may also report A company you are familiar with.In this case, you might be willing to pay much more than the 100-page copy of The Boston Globe for a 10-page, But the Boston Globe edited just for you.You will consume every bit of it.You can call it My Daily (The Daily Me). But, on Sunday afternoon, we like to read the newspaper in a more peaceful mood, learn about things we never knew we would be interested in, do crossword puzzles, read funny cartoons, and find great sales. advertise.This could be called "Our Daily" (TheDailyus).The last thing you want on a rainy Sunday afternoon is a nervous interface agent trying desperately to weed out seemingly irrelevant information for you. These are not two diametrically opposed states that are black and white.We tend to move between these two poles, wanting less or more personal information depending on how much time we have on hand, what time of day it is, and our mood.Imagine a news-reporting computer monitor with a knob on it. You can adjust the level of personalization of the news content like adjusting the volume.You can have a number of different controls, including one that slides left or right, allowing you to adjust the politics of a story (left or right) when reading a story about public affairs. In doing so, the controls change the window in which you view the news, changing the size of the window and the style in which it presents the news.In the distant future, interface agents will read, listen, and watch the full picture of every news item.And in the near future, this filtering process will be done with the help of message headers (that is, bits about bits).Trusted digital relative In the US, TVGuide's profits actually exceed those of all four networks combined. What it represents is that information about information can be more valuable than information itself.When we think about new ways to send information, our minds are stuck with the idea of ​​"scrolling through information" and "changing channels back and forth" that don't work anymore.When we have 1,000 channels, if you jump from one station to another, and only stay on each station for 3 seconds, it will take you almost an hour to scan all the channels from beginning to end. Before you can judge which show is the funniest, the show is over. When I want to go out and see a movie, I don't choose movies by reading reviews, I ask my sister-in-law for her opinion.We all have relatives and friends like this who know the movie well and know us well.All we need now is a digital relative. In fact, this concept of "agent" embodied in helping others often combines professional knowledge with understanding of you.A good travel agent combines what they know about hotels and restaurants with what they know about you (clues often come from your impressions of other hotels and restaurants).The real estate agent will deduce your preferred model of home from a series of houses that more or less meet your taste.Now imagine a telephone answering agent, press agent or e-mail manager!What they all have in common is the ability to mimic the way you do things. It's not as simple as filling out a survey or having a firm grip on you.Interface agents must also learn and grow like human friends and assistants.This is also easier said than done.Only recently have we had a glimpse of how to get brain models to learn about people. When I talk about interface agents, I'm often asked, "Did you mean artificial intelligence?" The answer is "yes." But there is a slight skepticism in this question, mainly because of the many false hopes and high promises artificial intelligence has given people in the past.In addition, many people remain deeply disturbed by the notion that machines can be intelligent. It is generally accepted that Alan Turing (A1an Turing) first seriously proposed the concept of machine intelligence in his 1950 paper "Computing Machinery and Intelligence".Later, pioneers such as Marvin Minsky continued to conduct in-depth discussions on the research of pure artificial intelligence.They ask themselves many questions about how to recognize text, understand emotion, appreciate humor, and deduce one set of metaphors from another.For example, after the series of English letters 0, T, T, F, F, according to the internal logic, which letters should be followed? Around 1975, when computing resources began to have the ability to solve intuitive problems and exhibit intelligent behavior, artificial intelligence research slumped.Scientists at that time chose to study easy-to-achieve and marketable application technologies such as robotics (such as securities trading and civil aviation reservation systems), so the more profound and fundamental problems of artificial intelligence and learning were ignored. Minsky is quick to point out that even though today's computers are exceptionally good at grasping airline reservations (a thing that's almost beyond logical systems), they still can't exhibit the common sense of a three- or four-year-old. They cannot tell the difference between a cat and a dog.Subjects like common sense have now moved from the backstage of scientific research to center stage.This is very important, because interface agents without common sense can feel like a thorn in your back. By the way, the answer to the question about connecting letters mentioned above should be S.This order comes from the first digit of each word in the English number arrangement one (one), two (two), three (three), four (four), five (five), six (six), seven (seven)... letters (0, T, T, F, F, S, S).From Centralization to Decentralization Many people tend to think of the interface agent of the future as the centralized, all-knowing machine described by the novelist George Orwell.Rather, what is more likely is a combination of many computer programs and personal tools, each of which is good at doing a certain kind of thing and communicating with other programs.This image is a copy of Minsky's 1987 book The Society of Mind.In this book, he pointed out that intelligence does not exist in the central processing unit, but in the collective behavior of many special-purpose machines that are closely connected to each other. This view shatters many past stereotypes.In his 1994 book, Turtles, Termites, and Traffic Jams, Mitchell Resnick called this stereotype "a totalitarian mindset."We are intensively trained to attribute complex phenomena to the manipulation of an agent.For example, we usually think that the first bird in the "herringbone"-shaped flock is just the head bird, and the rest of the birds just follow the leader.But in fact, it's not.Order is formed as a collective result of individual behaviors of birds that are highly responsive to each other.The flock of birds just followed simple rules of harmony, and no single bird was in the middle directing the overall situation.To illustrate his point, Resnick also created situations that surprised many to find themselves caught in the same process. I recently experienced Resnick's demonstration first-hand in the Great Hall of the Massachusetts Institute of Technology.There were about 1,200 people in the audience.Resnick asked everyone to start applauding, and the applause should be as coordinated as possible.As a result, in the absence of Resnick's command at all, the entire auditorium was filled with rhythmic applause in less than 2 seconds.You might as well try it out for yourself, even with a much smaller number of people, the results are still jaw-dropping.The stunned reaction of the audience illustrates how little we know about the coordination that emerges from the actions of independent individuals. This is not to say that your scheduling agent will therefore schedule the meeting without consulting your travel agent.Rather, not all information exchanges and decisions require the approval of a central authority. This approach may not be suitable for civil aviation reservation systems, but it is increasingly seen as a viable method of managing organizations and governments.The higher the degree of internal communication and power decentralization within a structure, the stronger its adaptability and survivability, and it will surely be able to survive and develop more sustainably. For a long time, the concept of decentralization has been lauded, but in practice, it has been difficult to move forward.The Internet provides global access to communication, free from any censorship, and is therefore particularly prevalent in places like Singapore, where press freedom is rare but the Internet is ubiquitous. Interface agents, like information and organizations, will gradually move towards a decentralized structure.Just like an army commander sends scouts out to scout the way or a county sheriff sends out a team of security guards, you send agents to gather information for you.The agent will then assign an agent.Such a deduction layer by layer.But don't forget how this process starts: you delegate your request to the interface, instead of diving into the WorldWideWeb (WorldWideWeb) to find it. This future model is very different from interface design that adds a human element.The look and feel of an interface is important, but it's nothing compared to the intelligence.In fact, the most common form of interface in the future will be a small hole or two in plastic or metal with a small microphone inside to pick up your voice. It's also important to recognize that there's a big difference between the idea of ​​an interface agent and the current Internet frenzy and the way people browse the Internet with Mosaic.Hackers can surf this new medium, explore the ocean of knowledge, and indulge in all kinds of new ways to socialize.This worldwide hot and cold internet fever won't abate or fade, but it's just one type of behavior, more like direct manipulation than delegated agency. Our interface will appear of all kinds.Because everyone has different information preferences, entertainment habits, and social behaviors, your interface will be different from mine.Everyone takes what they need from the huge palette of digital life.