# §Augmented Knowledge Generation Augmented Knowledge Generation is an umbrella for the various lines of inquiry to use new technology (mostly computers, but could also be process technology) to unlock new knowledge that people wouldn’t come up with “the old fashioned way.” I think it’s worthwhile to put all of them in the same place because if you’re purely interested in “[[§How do we get more awesome sci-fi shit?]]” it’s possible the most useful approach is a hybrid. * [[Augmented knowledge generation sits at the intersection of several “named” disciplines]] * [[Information retrieval is the study of generically getting information out of a collection of resources.]] * [[Text mining]] * [[Knowledge Discovery and Data Mining - ToA]] * [[Literature-based discovery]] * [[Expertise Finding (wikipedia)]] * [[Crowd-based knowledge synthesis - ToA]] * [[§Knowledge Graphs]] At high level, all of the approaches use some combination of * [[Algorithmic knowledge generation]] where a computer does some of the legwork of creating new knowledge. * [[Collective Intelligence]] where you use new technology and processes to enable better coordination between people and enable people to contribute to knowledge generation who would not otherwise be able to. I would call these “techniques” rather than “approaches.” * [[Crowd-based knowledge synthesis - ToA]] * [[Designed Serendipity]] ## List of Different Approaches * [[Undiscovered public knowledge discovery (UPKD)]] * [[Using published literature to find causal relationships that can be connected to lead to a new causal relationship]] * [[Using computers to search databases of existing molecules or compounds to suggest new uses]] ### History * [[bushWeMayThink1945]] * [[Undiscovered Public Knowledge (Paper)]] * [[DARPA and IARPA have run at least five programs focused on UPKD covering both crowd and algorithmic knowledge discovery]] ### Criticism of UPKD * [[Most human knowledge is not encoded]] * [[UPKD requires public knowledge to be both discoverable and interpretable in order to produce useful results]] * [[In order for knowledge to be discoverable and interpretable that means that it needs to be encodable]] * [[Of conception and implementation of undiscovered public knowledge, conception is the low-hanging fruit]] * [[Just having the piece of literature doesn’t solve the problem]] * [[The limiting reagent for UPKD is better methods for increasing interpretability]] * [[People have tried many different attempts to get experts to encode public knowledge in ways that make it easier to connect the pieces]] * [[Structuring knowledge is expensive]] * [[Commercial approaches to UPKD act like specialized search engines]] * [[Knowledge discovery is just not that useful]] ### Other Methods * [[Training computers on existing molecules or compounds and using them to predict properties of new molecules or compounds]] * [[Using computers to simulate possible experiment targets is still primitive]] * [[Using crowdsourced non-experts to perform optimizations]] * Good for [[Problems where it is easy to know whether an answer is correct but hard to get to that answer]] * [[High Throughput Experimentation - HTE]]/[[Using robotics to carry out automated experiments]] * [[Designed Serendipity]] ### References * [[Michael Nielsen characterizes institution building as making previously illegible things legible]]’s [[Reinventing Discovery]] presents a compelling vision, challenges to, and state of Augmented Knowledge Generation in 2012 * [[Connor Coley]] ’s [[Autonomous Discovery in the Chemical Sciences I + II]] and [[Accelerating Advanced Energy Materials Discovery by Integrating High-Throughput Methods with Artificial Intelligence]] give good overviews of Augmented knowledge generation in chemistry and materials. [Web URL for this note](http://notes.benjaminreinhardt.com/§Augmented+Knowledge+Generation) [Comment on this note](http://via.hypothes.is/http://notes.benjaminreinhardt.com/§Augmented+Knowledge+Generation)