Andrew McAfee Keynote: Our Computer Overlords #e2conf

Andrew McAfee is the principal research scientist at MIT.

[These are my notes from the Enterprise 2.0 Conference 2012 in Boston. Since I’m publishing them as soon as possible after the end of a session, they may contain the occasional typographical or grammatical error. Please excuse those. To the extent I’ve made any editorial comments, I’ve shown those in brackets.]


  • The World is One Big Data ProblemThis has become an article of faith in silicon valley. A lot of start-ups are now exclusively focused on cracking the big data challenge. According to Google, they don’t have better algorithms than the rest of the world — they just have better data.
  • Computers are Getting Better IBM has produced a graph that shows how Watson has improved itself with respect to percent of questions answered and precision. Relentlessly, month by month, Watson got better. This was just one domain that computers mastered. Even very human functions such as writing prose can be done reasonably well. (McAfee showed a screenshot showing an article in Forbes regarding Apple. Interestingly, it was written by a computer based on algorithms. Narrative Science is the company that produced it.) Long-form journalism may be safe from computers for the time being, but the more anodyne wwriting is completely within the realm of the possible for computers. Even pathology is being mastered by computers. Once they built the relevant algorithms, they discovered that the computer was at least as good as a human pathologist at detecting breast cancer. Further, the computer was able to identify three things in the samples that humans were not yet trained to observe.
  • Is the Physical World Safe from Computers? It is conceivable that computer algorithms could master knowledge work, but can they master the physical world? Google’s autonomous car actually drives itself. McAfee and his co-author drove it down Route 101 in stop-and-go traffic. The computerized car did just fine.
  • Humans Were Never All That Good In every realm checked, we’re discovering that our mastery nearly isn’t as complete as we supposed. Take for example, the art of predicting good bordeaux wine. A computer programmer was able to create a program that could reliably predict which year’s bordeaux production would be best, based on agricultural data. In the legal domain, a computer beat the best legal minds at predicting Supreme Court decisions. With respect purchasing decisions, a fairly simple model will allow a computer do this better than humans with experience and expert judgment. In 136 man-vs-machine studies, in only 8 cases did the humans outperform the computers. The researchers attributed these 8 to either computational error or the fact that the computers lacked sufficient data. This problem will go away soon.
  • What Bad Strategies Could We Adopt? (1) Ignore the problem until the competition adopts the computers; (2) be a Luddite and destroy the machines.
  • Race with the Machines McAfee believes that partnering with computers is the best strategy for us to adopt. Deloitte’s Center for the Edge study shows that quality improved as a wiki as edited more. Further, Lynn Wu discovered that people who participate in Enterprise 2.0 social networks tend to survive rounds of layoffs.
  • I for one welcome our Computer Overlords. What about You?

4 thoughts on “Andrew McAfee Keynote: Our Computer Overlords #e2conf

  1. I second the thanks. Am a big fan of the work of McAfee (& his co-author Eric Brynjolfsson) and it’s nice to get more of his content.

Comments are closed.

Create a website or blog at

Up ↑