From 0 to 1: unlocking the secrets of business and the future

Chapter 20 Humans and Machines

As mature industries are increasingly depressed, information technology has become synonymous with technology due to its rapid development.More than 15 billion people use portable devices to access timely information.Our smartphones are now a thousand times faster than the computers that guided astronauts to the moon.If Moore's Law continues to apply, future computers will be even more powerful.

Computers are also capable enough to beat humans in activities we once thought were the domain of humans. In 1997, IBM's Deep Blue defeated world chess champion Garry Kasparov. In 2011, top contestant Ken Jennings on the quiz show Jeopardy! lost to IBM's Watson computer.Google's self-driving cars are already driving on California roads.Although the well-known racing driver Dale Earnhardt Jr. does not feel threatened by machines, the British "Guardian" worries for millions of employed drivers and taxi drivers around the world that self-driving cars "will bring the next wave of unemployment".

Everyone expects that computers will be able to do more things in the future, so many people wonder: 30 years later, will human beings still have anything to do?Venture capitalist Mark Anderson categorically declared that “software is eating the world.” And his colleague Andy Kessler seemed gleeful when he explained that the best way to increase productivity was to “get rid of humans.” Forbes, on the other hand, seemed worried, asking its readers the following question: Will machines replace you?

Futurists seem to expect an affirmative answer.Luddites [Luddites are skilled workers who were unemployed because machines replaced manpower during the British Industrial Revolution in the 19th century.It is now extended to people who hold anti-mechanization and anti-automation views. ——Editor's Note] is afraid of being replaced by machines, and would rather stop developing new technologies altogether.Neither side disputes the premise that good computers will necessarily replace humans.However, this premise is false: computers are tools to complement humans, not substitutes.In the next few decades, the most valuable industries will still be established by entrepreneurs who develop computers to enhance human capabilities, not eliminate them.

Globalization vs. Technology
Fifteen years ago, Americans worried about competition from cheap Mexican labor, which made sense because anyone could be replaced by someone else.Today one hears again what Ross Perot called the "big sucking sound," this time traced not to cheap factories in Tijuana but to server farms in Texas.Americans are afraid of technologies that will appear in the not-too-distant future because they see it as a repeat of the globalization of the past few years.But things are very different now: people compete for jobs and resources, computers don't.

Globalization means replacing
While Perot warned against foreign competition, Bush and Bill Clinton were advocating free trade: Everyone has expertise in their jobs, so in theory, as long as people develop their expertise to their own advantage and use it to mutually trade, the economy maximizes wealth.But in practice, at least for many workers, the benefits of free trade are not so obvious.The benefits of trade are greatest when the relative advantage is very different, but the world is full of people willing to do low-paying repetitive jobs.

People compete not only for jobs but for the same resources.While Americans are buying toys and textiles imported from China at low prices, they need to pay more for gasoline as millions of Chinese car owners start to compete for gasoline.Whether it's eating shark fin in Shanghai or fish tacos in San Diego, people need food and they need housing.And people aren't just satisfied with food and clothing -- as globalization advances, people's needs will continue to increase.Now that hundreds of millions of Chinese peasants have finally reached food and clothing levels, they naturally want to eat more meat and less grain.The desires of the upper classes are even more strikingly consistent: oligarchs from St. Petersburg to Pyongyang love crystal champagne.

tech means supplement
Now think about the prospect of competition from computers rather than humans.On the supply side, the difference between computers and humans is much greater than the difference between people: there are fundamental differences in what humans and machines are good at.Human beings are conscious and good at making plans and making decisions in complex situations, but not good at processing large amounts of data.Computers are just the opposite. They are good at efficient data processing, but they cannot make basic judgments that humans can easily make.

To get an idea of ​​the extent of the difference, let's look at another of Google's human-machine replacement projects. In 2012, a Google supercomputer made headlines when it was able to identify cats with 1000 percent accuracy after scanning 75 million YouTube video thumbnails.It may seem incredible—but a 4-year-old can do it with ease.The cheapest laptop can beat the smartest mathematician, but even a supercomputer with 16000 CPUs can't compete with a child in other ways, and it's not just a matter of one being more powerful than the other, but are fundamentally different.

The stark difference between humans and machines means that the outcomes of working with a computer are far greater than those of trading with a human.We do not trade with computers any more than we trade with livestock or lamps.The point is this: Computers are tools, not competitors.

The difference is even greater on the demand side.Unlike people in industrialized countries, computers don't aspire to lavish meals or sea-view villas in Cap-Ferrat, France;We develop new computer technologies to solve problems, which means we have a super-specialized partner who can serve us efficiently without competing with us for resources.Rather, technology is the only way to escape competition in this globalized world.Although computers are getting more powerful, they cannot replace humans: they only complement them.

Human-computer complementarity for enterprises
The complementarity of humans and computers is not just a macro fact, but a way to create great things.My experience at PayPal taught me this. In mid-2000, we managed to survive the dot-com bankruptcy and grow rapidly, but still faced a huge problem: losing tens of millions of dollars every month to credit card fraud.Hundreds of thousands of transactions are processed every minute, so it is impossible to check them one by one - no quality control team can achieve this kind of speed.

So we did what any team of engineers would do: we used automation to find solutions.First, Max Levchin assembled an elite team of mathematicians to scrutinize fraudulent transactions.Then use the research findings to write automatic identification software to cancel fraudulent transactions in real time.But this measure quickly became ineffective, as after an hour or two, the thieves found out and changed tactics.Our adversaries are adaptable, while our software is slow to react.

Scammers eluded our automated detection algorithms, but we found that they could not easily fool human analysts.So Marks led engineers to rewrite the software with a mixed strategy: the program flags suspicious transactions on a well-designed user interface, and then manually reviews their legality.Thanks to this hybrid system, we caught the Russian thief who boasted that he was invincible, so we gave this system a Russian name - "Igor".And, with this system in place, we turned a profit in the first quarter of 2002 (compared to a loss of $2001 million per quarter in 2930).The FBI came to us asking if we would lend Igor to help them detect financial crimes.This led Max to call himself "Sherlock Holmes the Internet spy," and he was.

This combination of man and machine has allowed PayPal to gain a foothold in the business world, and hundreds of small merchants are willing to accept payments online to grow.None of these results would be possible without a human-machine solution—even though most people know nothing about it.

After selling PayPal in 2002, I still worked hard on the combination of man and machine: the combination of man and machine is more effective than fighting alone, so what valuable business can be built on this core foundation?In the second year, my old Stanford classmate Alex Karp and software engineer Stephen Cohen and I came up with the idea of ​​starting a company: using the human-computer composite model of PayPal's security authentication system to identify terrorists and financial fraud.We knew the FBI was interested, so in 2004 we co-founded Palantir, a software company that helps people extract useful information from different sources. By 2014, Palantir’s sales to $10 billion. Forbes called Palantir's software "killerware" because it was rumored to have helped the U.S. government find Osama bin Laden.

I can't comment on the operational details, but it's safe to say that human intelligence or computers alone are not enough to keep us safe.The two largest intelligence agencies in the United States use very different methods: The CIA tends to use people, while the NSA tends to use computers.CIA analysts have too much noise to filter out to identify serious threats.The computers of the National Security Agency are very capable of processing data, but the machines themselves cannot identify whether someone is planning a terrorist act.Palantir is working to overcome both of these flaws: Using Palantir's software to analyze government-provided data (such as call logs of extremist clerics in Yemen, bank accounts linked to terrorism) and flag suspicious activity, Available for review by trained analysts.

In addition to helping find terrorists, using Palantir's software, analysts can predict where insurgents in Afghanistan will plant explosive devices; prosecute high-profile insider trading cases; fight the world's largest child pornography ring; support disease control and prevention Center to fight outbreaks of foodborne disease; through advanced fraud detection software, can save commercial banks and governments hundreds of millions of dollars in annual losses.

Advanced software makes this possible, but more importantly, human analysts, prosecutors, scientists, financial experts, without their active participation, software is useless.

Consider what experts do today.Lawyers must speak differently about solutions to difficult problems—changing their rhetoric according to who they are talking to—clients, opposing counsel, judges, and so on.Doctors must have the ability to communicate diagnosis and treatment results with ordinary patients who are not experts.Good teachers are not just proficient in the subjects they teach, they must also understand how to adapt their teaching methods to the interests and learning styles of their students.Computers may be able to perform some tasks, but they cannot be integrated efficiently.Advanced technologies in the legal, medical, and educational fields cannot replace experts, but can only help them do better.

That's exactly what LinkedIn helps recruiters do.When LinkedIn launched in 2003, it neither consulted recruiters to find areas for improvement nor wrote software to completely replace recruiters.Recruiting is half detective work and half salesmanship, with recruiters scrutinizing candidate histories, assessing motivation and adaptability, and convincing the best talent to join the team.It’s impossible for a computer to do everything efficiently, so LinkedIn started by changing the way recruiters work.More than 97% of recruiters now use LinkedIn's network to screen candidates using its powerful search filters, and the network creates value for the hundreds of millions of professionals who use it to manage their personal brands.If LinkedIn had simply replaced recruiters with technology, it wouldn't be the size it is today.

Awareness of computer science

Why do so many people ignore the complementary power of computers?This starts with school education.Software engineers work on developing projects that replace human labor, that's what they do.Academics make a name for themselves through professional research, and their main goal is to publish, and publishing means respecting the boundaries of a particular discipline.For computer scientists, it means reducing the functions of humans to specific tasks, and computers can be trained to complete each task one by one.

At the forefront of computer science today, the term "machine learning capability" has inspired the fantasy of machines replacing humans, and its advocates seem to believe that as long as enough training data is fed, computers can perform any task.Users of Netflix (Netflix) and Amazon have experienced the effects of computer learning first-hand: Both companies use specific algorithms to recommend products based on consumers' browsing and purchasing histories.The more data entered, the better the recommendations.The same is true for Google Translate. It supports translation in 80 languages. Although it is rough, it is barely usable. This is not because the software understands the language, but because it can perform statistical analysis on the text of a huge corpus and extract sentence patterns.

Another buzzword that embodies the tendency for machines to replace humans is "big data."Companies today are obsessed with data, under the false belief that more data creates more value.But big data is usually silent data. Computers can find patterns that humans do not notice, but they cannot compare patterns compiled from different sources, nor can they use these data to explain complex human behavior.Actionable insights can only be given by human analysts (or the kind of artificial intelligence that only exists in science fiction).

We are obsessed with big data simply because we think technology is weird.We marvel at some small achievements made by computers alone, but we ignore the huge progress made by humans with the assistance of computers, because human participation dilutes its mystery.Although Watson, Deep Blue Computer and increasingly powerful algorithms are cool, the most valuable companies in the future will definitely not rely on computers to solve problems alone, but focus on how computers can help humans solve difficult problems.

Smart Computers: Foe or Friend?

The future of computing is full of unknowns.The ever-sophisticated robot intelligence that heralds future trends like Siri and Watson is gaining popularity; once computers can answer all of our questions, they may ask why they have to give in at all for us?
The logical end point of alternative thinking is "powerful artificial intelligence": computers eclipsing humans in every important domain.Of course, Luddites are horrified by the possibility.This has even unsettled futurists, as it is uncertain whether powerful artificial intelligence will save or destroy humanity.Technology is supposed to increase man's control over nature and reduce the contingency of human life; building computers that are so smart must have both pros and cons.Strong artificial intelligence is like a cosmic lottery: we win and get Utopia; we lose and get replaced by Skynet.

But even if strong artificial intelligence is not an unpredictable mystery but a real possibility, that time will not come soon: being replaced by computers is a problem that human beings should worry about in the 22nd century.Fear of uncertainty about the distant future should not prevent us from making definite plans now.Luddites believe that we shouldn't be building computers that could one day replace humans, while fanatical futurists hold the opposite view.These two views are mutually exclusive, but not all views: there is a huge space between these two extremes, and sane people can build a good world in the coming decades.Our innovations in the use of computers will not only help humans do what they already do, but also do things that were previously unimaginable.

(End of this chapter)