Search interesting materials

Monday, January 16, 2012

The new world of computers

I read a beautiful article, Why software is eating the world by Marc Andreessen. It made me reflect on how the world of computers and networks has evolved over the last 20 years. It is comfortable for us to think that the world has only evolved in incremental ways. But as I look around me, it seems that the hard-driving pace of change on many fronts has added up to fundamental change, to places far away from the comfort zone of people of my vintage.

All the way till the 1980s, business computing was dominated by databases. The basic story was one of capturing data, storing it, and summoning it forth with queries.

At first, databases were the exclusive preserve of mainframes and minicomputer. The PC revolution made it possible for small databases to be held on the desktop. It's interesting to note that at first, we got PCs without networks. We evolved from databases stored on remote mainframes or minicomputers to databases stored on PCs. All the way into the late 1980s, it was quite a cool thing to have a standalone PC holding a database where certain queries could be executed.

The first wave of change, of the early 1990s, was networks in the form of TCP/IP (the universal communication protocol) and the Internet (the universal network). Now, suddenly, the data centre became more interesting. Instead of storing and manipulating data at the desktop, we could do so many better things by storing and manipulating data at a big central computer. The desktop diminished from being the location of data and computation to being the location of the user interaction.

Then came a series of surprises which have added up to a qualitative shift.

1. The network got ubiquitous

First, the Internet went everywhere for the road warrior armed with the laptop computer. Crashing prices of laptop computers and then netbooks meant that essentially everyone had one. So workers started spending much more time outside the office (with 100% connectivity).

Software had to adapt itself to reach out on an Internet scale. This killed off applications which worked on the scale of the LAN. The software that the busy road warrier used was the software that worked effortlessly on his laptop.

Today, 1 Mb/s wireless networks are common and 50 to 100 Mb/s offerings are on the anvil. This is relentlessly shifting the balance of convenience to mobility.

In a place like India, the low-end staff might not have netbooks and/or Internet on the go. So for certain very low-end applications, it might make sense to hug the desktop at the workplace. For any modestly well paid person, laptops / netbooks coupled with 3g or CDMA networks are the norm, and hence being tethered to the office network is quite limiting.

2. The user interfaces got better

In the 1980s, software came with fat manuals. Users actually sat down in training classes. A remarkable feature of the new world is how the manuals and training are gone. Software is incredibly capable but there are no manuals. Google maps or Amazon or Apple Mail are very powerful programs, but the fundamental assumption is that a reasonable person can just start tinkering with them and learn more as he goes.

The modern office worker gets no formal training in software all his life. The modern knowledge worker learns major tools (e.g. a programming language) and often puts in enormous effort for these. But for the rest, the ordinary flow of day to day life, where new software systems come up all the time, is done without formal training.

Once the modern office worker faces high quality UI design from google and such like, where there is zero training and zero marketing, it became much harder to accept training. Standards have changed; in the olden days, people would actually try to learn. Today, knowledge workers are willing to get training in programming languages (e.g. R or Stata) but not in applications. The MBAs are generally training-proof.

3. All of us got busy

There was a time when one purposedly went about the work day systematically doing certain things with certain software tools. Knowledge workers have become deluged with information and with stimuli. We have gone from being an information scarce economy to being an attention scarce economy.

Software and information systems are now competing for the attention of the user. The scarce resource is now the mind share of the user. This is linked to the problem of user interfaces. If something has a complicated user interface, and there are a hundred other tasks that need to be done, the user ignores the complicated thing. Software systems that don't fly immediately just die.

4. Peers determine where attention is directed

In a world where the knowledge worker is bombarded with hundreds of things every day, what does he do? He tends to direct his scarce time into the things that come well recommended. The recommendations of respected peers are supremely important in determining what a person does.

High powered sales compaigns have lost power. The person just asks his friends what they do. The impulses through the day coming into each person - over email, IM, twitter, social networks, etc. - are the de facto controllers of the persons' time.

Peers are thus the gatekeepers to the user. The stuff that is striking and remarkable gets noticed and pointed to friends. What gets pointed tends to get a high google pagerank.

The importance of high pressure sales dropped. Some of the most successful firms got by with negligible sales departments. Their stuff was intuitive and good, and got immediately picked up.

5. Network effects leading to user generated content

The old model was one of corporations producing information and users consuming information. In that power structure, the user was only a source of revenue.

In the new world, the critical story is about kicking off network effects. The systems that win are those that get better because of one more user interaction.

At the simplest, user interactions kick off impulses to peers which brings in more customers (viral marketing). But very soon, user interactions generate relevant data. Google watches what users click and uses that to improve search. Amazon tells you that the people who liked this book also liked that book. Amazon has user-generated content in terms of reviews.

Good systems create a warm and supportive environment in which users contribute bug fixes, feature suggestions. These systems ride the power of user eyeballs and brains to get better. The power structure has changed. IBM DB2 used to be designed in a temple and then went out to the helpess masses. Google's world is critically linked to the users at so many levels (a receptive environment for bug reports, feature requests, user generated content, and usage data being turned back into strengthening the system).

The bottom line: Successful designers found ways to harness every single user and user interaction to build the quality, the content and the footprint of the system. Stalinist structures, which disempowered the user and treated him only as a source of revenue, stand isolated and stagnant.

6. Loss of power of enterprise IT

In the old world, enterprise IT mattered more. Grave decisions were made by enterprise IT managers and then thousands of users fell in line. In the new world, users forge ahead with their laptops and tablets and mobile phones, exercising enormous autonomous choice about how they spend their time. Consumer considerations, and the loyalty of each individual user, are far more important than they used to be. The enterprise IT department is much less of a gatekeeper. In the olden days, hardware and software was sold to enterprise IT, which made decisions for everyone inside the organisation. In the new world, usage is won one user at a time, and it is contestable every day.

7. CPUs became too cheap to meter

In the old world, computation was something scarce. The money that went into building data centres was carefully weighed. System designers carefully did things that were parsimonious in the use of CPU.

With the rise of parallel computation, bringing 1000 CPUs into a problem became cheap. Successful designers were those that found ways to deploy incredibly large amounts of computer power to do things that delight users. Google and amazon are spending millions of clock cycles in the back end, thinking about how to handle the next move, as the mouse cursor moves! When faced with a choice between doing something nice that users will like, versus doing something that saves compute power, the former always won.

8. Unexpected revenue sources

Who would have imagined that an ad agency would become the most powerful author of operating systems for mobile phones in the world? When hardware got dramatically cheap, and the Internet generated access to eyeballs on an unimaginable scale, new revenue models came about which were surprisingly different from the way we used to think earlier.


  1. I disagree with point 4. I heavily use review websites before making a decision on which product to buy ...the product can be electronics or car...But I always consider opinions of review websites.

  2. Awesome documentation of the evolution, sir. Thank you very much.

    I initially felt that cloud computing was a jargon and will not affect day-to-day life much. Mobile gadgets, tablets is leading to enterprise IT support having a policy of bring your own device to work and support it yourself. Probably desktops, servers, mainframes as we know will be history.

  3. From a public policy point of view, it is not clear if all this is completely good without simultaneous changes in the way we perceive "work", "productivity" etc.

    Software is eating jobs and every such technological change produces fewer net jobs.

    It's time our society, delivers the promised extra leisure time from all this automation. Without reducing human productivity, it would be impossible to provide paying jobs for everyone. Until now, we have pushed this issue away by allowing people to load up on debt or live on welfare of all kinds.


Please note: Comments are moderated. Only civilised conversation is permitted on this blog. Criticism is perfectly okay; uncivilised language is not. We delete any comment which is spam, has personal attacks against anyone, or uses foul language. We delete any comment which does not contribute to the intellectual discussion about the blog article in question.

LaTeX mathematics works. This means that if you want to say $10 you have to say \$10.