So far, following points have come up in the discussions concerning EM technology:
1. Technological Focus
It makes no sense to re-invent the wheel. EM tech should rather probably tap into existent systems. API’s from Facebook, Twitter, Evernote, Google, Wikipedia and so forth could be used for the social, search and storage functionality needed, respectively.
Initial focal points for EM tech could be as follows:
1) Back end: a graph database for the semantic engine, driven by semantic algorithms collected or created to drive a powerful semantic tagging system to sort the gathered information.
2) Front end: a network-based (touch) UI with which the user could intuitively manipulate massive amounts of data.
2. Semantical Considerations
Semantics seems quite central to EM tech. The important things to consider are:
1) how to access information fast
2) how to store information efficiently, and
3) how to determine the credibility of information sources.
The first two items are addressed in creating a simple but powerful semantic engine that can determine a tagging hierarchy for the stored information. The third one involves some kind of a social evaluation system.
The semantics would function by mining data from the note/social stream. It could further be factored by existent tags, and a semantic structure guided by the storing guidelines of the user (i.e. which folder she decides to put the info). And, the coolest bit, we could return top-level tags by a simple combination of intensional semantics and Olli’s network analytics.
By combining aspects from intensional semantics with network analytics, it seems one could build a tagging and sorting model that draws the intensional weights of significant terms from the social network, and furthermore returns top-level terms where needed. By adding to this a simple social learning functionality à la StumbleUpon, one can pretty quickly make a working point-and-shoot archiving system for any information put into the system.
Furthermore, the social network itself can be configured to be less homogenic by assigning credibility values to users.
3. Credibility Values
What if we were to combine assigning credibility values (i.e. who’s got the most credible info on shoes, philosophy or the Simpsons) with a social game-like dynamic?
In other words, how about if users could assign credibility ratings to friends in a social network. For example, I could assign Petri as a credible source for gaming information, Olli as a credible network guy and Timo T. as a credible philosophy guy. These could then play in as factors in determining the overall semantic load in the network: the way I or Timo T. tag our philosophy-related notes is likely to be much more accurate than the way we tag our, say, Desperate Housewives -related info :). This could be driven by some kind of status or reward based system, such as is implemented in FourSquare’s assignment of mayorships.
4. Development Phases
It seems the tech would be best built in phases, rather than going full-salvo EM all at once. It would be nice to start with a simple angle that would be immediately approachable and needed by the basic user, and then expand the EM / Singularity aspects from there. Some tentative suggestions for the progression:
PHASE I: Social Memory
Could this be interesting enough to get people initially on the train?
This could work something like Memolane, except with semantics and with a more EM-focused approach. The basics of the network UI could be implemented here already, and the basic semantic architecture could be in place when launching. This phase would effectively produce an interface to access everything that’s in your social network as a full-text searchable, sematically taggable database.
PHASE II: Personal Memory
Once the semantic engine and the UI are in place, one could either tap into some storage API to create a personal memory extension of the above. Now, in addition to the social memory automatically generated by Facebook, Twitter, etc., one could have a personal memory akin to Evernote, except with all the tagging and organizing goodies produced by the semantic and social engine. (This is actually pretty close to what it seems SpringPad is already aiming at.)
Basically, this could be something like a point-and-shoot UI for Evernote that generates automatically an intuitive tag hierarchy for accessing notes later.
PHASE III: Collective Memory
Next phase would be to expand the above by integrating search engine API’s. This would produce an easy access to data, once again semantically sortable and taggable, and displayed as a network rather than as a list.
The semantic and UI benefits of the above developments could thus be ported also to looking up data on the internet, speeding up and focusing access.
PHASE IV: Social Singularity
At this phase we would be looking at something like a combined social-evernote-google -system in place, where one can tap into precisely that information that is relevant, as judged by the mined information, the semantic weights determined by the social network, and the credibility assigned to individual users.
Now, it should be no problem adding to this a realtime q&a service like Aardvark or Quora. You would already have your experts determined in the network, so directing questions should be no problem. The social network could be Facebook, so getting to critical mass could actually be simpler than we thought before. It’s possible that getting from collective memory to social singularity would simply require adding a real-time q&a plugin or some such thing.