Sign In

Communications of the ACM

BLOG@CACM

The Chaos of the Internet as an External Brain; and More


View as: Print Mobile App ACM Digital Library Full Text (PDF) In the Digital Edition Share: Send by email Share on reddit Share on StumbleUpon Share on Hacker News Share on Tweeter Share on Facebook
BLOG@CACM logo

Greg Linden's "The Rise of the External Brain"

http://cacm.acm.org/blogs/blog-cacm/54333

From the early days of computers, people have speculated that computers would be used to supplement our intelligence. Extended stores of knowledge, memories once forgotten, computational feats, and expert advice would all be at our fingertips.

In the last decades, most of the work toward this dream has been in the form of trying to build artificial intelligence. By carefully encoding expert knowledge into a refined and well-pruned database, researchers strove to build a reliable assistant to help with tasks. Sadly, this effort was always thwarted by the complexity of the system and environment, with too many variables and uncertainty for any small team to fully anticipate.

Success now is coming from an entirely unexpected sourcethe chaos of the Internet. Google has become our external brain, sifting through the extended stores of knowledge offered by multitudes, helping us remember what we once found, and locating advice from people who have been where we now want to go.

For example, the other day I was trying to describe to someone how mitochondria oddly have a separate genome, but could not recall the details. A search for "mitochondria" yielded a Wikipedia page that refreshed my memory. Later, I was wondering if traveling by train or flying between Venice and Rome was a better choice; advice arrived immediately on a search for "train flying venice rome." Recently, I had forgotten the background of a colleague, which was restored again with a quick search on her name. Hundreds of times a day, I access this external brain, supplementing what is lost or incomplete in my own.

This external brain is not programmed with knowledge, at least not in the sense we expected. There is no system of rules, no encoding of experts, no logical reasoning. There is precious little understanding of information, at least not in the search itself. There is knowledge in the many voices that make up the data on the Web, but no synthesis of those voices.

Perhaps we should have expected this. Our brains, after all, are a controlled storm of competing patterns and signals, a mishmash of evolutionary agglomeration that is barely functional and easily fooled. From this chaos can come brilliance, but also superstition, illusion, and psychosis. While early studies of the brain envisioned it as a disciplined and orderly structure, deeper investigation has proved otherwise.

And so it is fitting that the biggest progress on building an external brain also comes from chaos. Search engines pick out the gems in a democratic sea of competing signals, helping us find the brilliance that we seek. Occasionally, our external brain leads us astray, as does our internal brain, but therein lies both the risk and beauty of building a brain on disorder.

Ed H. Chi's "The DARPA Network Challenge and the Design of Social Participation Systems"

http://cacm.acm.org/blogs/blog-cacm/60832

The DARPA Network Challenge recently made quite a splash across the Internet and the media. The task was to identify the exact location of 10 red weather balloons around the country. The winning team from MIT succeeded in identifying the locations of the balloons in less than nine hours.

There was recently a good article in Scientific American about the winning entry and the second-place team from Georgia Tech. The article details the way in which the teams tried to: (1) build social incentives into the system to get people to participate and to recommend their friends to participate; (2) how they managed to fight spam or noisy information from people trying to lead them astray. The MIT team, for example, required photo proofs of both the balloon and the official DARPA certificate of the balloon at each location, suggesting they realized that noisy or bad data is a real challenge in social participation systems.

But what did the challenge really teach us?

Looking back for the last decade or so, we have now gotten a taste of how mass-scale participation in social computing systems results in dramatic changes in the way science, government, health care, entertainment, and enterprises operate.

The primary issue relating to the design of social participation systems is understanding the relationship between usability, sociability, social capital, collective intelligence, and how to elicit effective action through design.

  • Usability concerns the ability for all users to contribute, regardless of their accessibility requirements and computing experience, and how to lower the interaction costs of working with social systems.
  • Sociability refers to the skill or tendency of being sociable and of interacting well with others. There is a huge role in how the designer can facilitate and lubricate social interactions amongst users of a system.
  • Social Capital refers to positions that people occupy in social networks, and their ability to utilize those positions for some goal. Designers need to enable people to sort themselves into comfortable positions in the social network, including leadership and follower positions.
  • Collective Intelligence (or Social Intelligence) refers to the emergence of intelligent behavior among groups of people. Designers can create mechanisms, such as voting systems, folksonomies, and other opinion aggregators, to ensure the emergence of social intelligence over time. (Note that the definition for "social intelligence" here differs from traditional use of the phrase in social psychology.)

The principal concern for designers of systems is to ensure the participants both give and get something from the system that is beneficial to both the individual as well as to the group. This may take the form of being challenged in their ideas, or to contribute to the overall knowledge of a domain, or to contribute their experiences of using a particular product or drug.

More importantly, social participation systems should encourage users to take part in effective action. One main design principle here is that effective action arises from collective action. That is, by encouraging participants to learn from each other and to form consensus, group goals will form, and action can be taken by the entire group.

The DARPA Network Challenge is interesting in that it was designed to see how we can get groups of people to take part in effective action. In that sense, the experiment was really quite successful. But we already have quite a good example of this in Wikipedia, in which a group of people came together to learn from each other's perspective, but they share a common goal to create an encyclopedia of the state of human knowledge for broader distribution. Here, collective action resulted in effective change in the way people access information.

Looking toward the next decade, the social computing research challenge is understanding how to replicate effective social actions in social participation systems, in domains such as health care, education, and open government. United, we might just solve some of the biggest problems in the world.

Mark Guzdial's "Are There Too Many IT Jobs or Too Many IT Workers?

http://cacm.acm.org/blogs/blog-cacm/67389

The latest U.S. Bureau of Labor Statistics (BLS) have been updated, as of November 2009, to reflect the Great Recession. The news is terrific for uscomputing is only forecast to grow, and at an amazing rate.

Via the Computing Community Consortium blog: "'Computer and mathematical' occupations are projected to grow by the largest percentage between now and 2018by 22.2%. In other words, 'Computer and mathematical' occupations are the fastest-growing occupational cluster within the fastest-growing major occupational group."

DARPA is so concerned about the lack of IT workers (and the lack of diversity among its workers) that it has launched a new research project to develop more and more diverse IT workers.

DARPA has launched a "far-out research" project to increase the number of students going into "CS-STEM" (computer science and science, technology, engineering, and mathematics) fields. Wired just covered this effort to address the "Geek shortage." What makes the Wired piece so interesting is the enormous and harsh pushback in the comments section, like the below:

"I'm 43, with a degree in software engineering and enjoy what I do for a living. But I wouldn't encourage my 12-year-old son to major in CS or similar because interesting, new project development jobs are the first to disappear in a down economy and non-cutting-edge skills are easily offshored and new hires are cheaper than retraining outdated workers."

"Why get a four-year degree for a career with a 15-year shelf life?"

Are these complaints from a vocal small group, or are do they represent a large constituency? Why is there this disconnect between claims of great need and claims of no jobs? Are old IT workers no longer what industry wants? Is BLS only counting newly created jobs and not steady-state jobs? Is the IT job market constantly churning? Is industry not training existing people and instead hiring new people? It's a real problem to argue for the need for more IT in the face of many (vocal) unemployed IT workers.

Back to Top

Authors

Greg Linden is the founder of Geeky Ventures.

Ed H. Chi is a research manager at Palo Alto Research Center.

Mark Guzdial is a professor at the Georgia Institute of Technology.

Back to Top

Footnotes

DOI: http://doi.acm.org/10.1145/1743546.1743551


©2010 ACM  0001-0782/10/0600  $10.00

Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee.

The Digital Library is published by the Association for Computing Machinery. Copyright © 2010 ACM, Inc.


 

No entries found