Part 1 of the dispatches from Defrag 2009 looked at new software developers using the metaphor of live updates for any cloud-based application, and whether that was a useful or limiting view. This time, I will riff off the wonderful presentation Khris Loux gave on the Synaptic Web to ponder the issue of whether long-term institutional “memories” and adequate data storage are being performed in a web environment largely defined by real-time services.
It’s useful to have a background in neural networks and cognitive neuroscience to grab the full import of what Loux had to say in his synaptic speech. If every node in a social network is like a synapse in a brain, then intelligence gains plasticity and adaptability by strengthening, parsing, or reinforcing synaptic connections. Loux gave an overview on what that might mean, but if you started pondering some of the neural-network work on back-propagation, genetic algorithms, and simulated annealing, you could quickly go stark raving mad considering the implications of how social networks might evolve.
But notice the comments from my last post about how environments like Twitter and Facebook are losing more and more long-term institutional memory. The web itself has mirroring, caching, and archived “dead sites,” to be sure, but the sheer volume of real-time streams makes the building of memories more problematic and exhausting. When I worked in the electronics and IT trade press, we always used to say that the day institutional memory broke down, was the day Stalinist re-interpretations of history would follow. To see how this could operate on a personal basis, check out the recent Doonesbury cartoon on optimizing your personal branding, then consider what happens without a detailed audit trail of tweets. The mind reels.
Plenty of web analytics companies are trying to build data management models for new social networking environments. One interesting example is Truviso. Michael Franklin, CTO of the startup that spun from UC Berkeley, said the company got started trying to structure data management for sensor networks, which are highly distributed, embedded nets based on wireless communications. When sensor nets failed to become as ubiquitous as we thought they’d be, Truviso founders realized their tools could be applied to rapidly changing decentralized networks like social networks. Once models are developed to monetize those “real-time streams,” I’m sure a business case for tools like Truviso’s will follow.
But let’s return to Loux’s synaptic analogy. The human brain reinforces synaptic connections for long-term memory by shifting to different oscillatory patterns during sleep cycles. Social networks know no equivalent of sleep. Cloud computing knows no distributed sleep state. Consequently, services and information evolve 24/7 with no synaptic reinforcement down-time. (No, Twitter and Facebook meltdowns don’t count.)
So as we shift from the neural-network analogy to the cloud analogy, we have to remember that clouds can be mighty ethereal in content, except when they burst forth with torrents. The mainstreaming of smartbooks and netbooks that depend on the cloud for all storage will escalate the messaging storms and tweetstorms we see on smartphones by a thousandfold. If we don’t work to build institutional memories in those clouds, they will dissipate before our eyes.
Loring
