Knowledge loss over time

Some of this could fit in with the Quantified Self, as well as Digital Preservation/Link Rot. Perhaps there is a better name for this that can be merged later. Data is being stored at a faster rate than ways to analyze all this data, but in the future, smarter machine learning techniques will make this easier.

this quote below (from is also apt:

The greatest waste of human potential is not transferring knowledge when people retire! WHAT A WASTE! All those years of knowledge and experience simply ‘put out to pasture’ when people retire!

I believe that people should be given a tax-break in their last 10 years of work if they can prove mentorship and knowledge transfer to newcomers. This would involve a managed program of sorts, but still I believe it is a travesty to see these people with years of experience and knowledge not passing it on.

Jonathan Blow documents some examples of knowledge loss in a talk. His examples are space travel (after moon landing until Elon Musk came along), Lycurgus Cup, Antikythera mechanism, and Greek fire (that’s the first ~20 minutes of the talk). The rest of the talk then argues that something similar is happening in software today, where everything is becoming more complicated, software is getting buggier over time, proliferation of intermediate layers of abstraction, dependence on this layer of abstraction (which means new people don’t know how to work with the lower levels anymore), etc.


AKC feels that this also happens with academia to a very large extent. Particularly when the best-taught courses are not archived (MOOCs are alleviating this, but this still has a long way to ago), and also when instructional methods have to be repeated between different students.

Also, people forget a shockingly high fraction of content they read, or “learn” from lecture (though this content is at least not permanently lost from the public domain). Most information communicated verbally is permanently lost and has little potential of being transmitted to other people.

In the far future, mind-uploading may be an answer (though one must also account the uncountable information one forgets as a child/teen/young adult)


How could we estimate how much utility is lost this way / how inefficient we’re being about it? Utility is lost when the cost of transferring old-people-information is lower than the value of that information. We see information being passed on as books, mentoring (how common?) and training / easing a new employee into the old position (other ways?). If this was really a problem among, say, personal assistants (such that personal assistants accrue useful information that would be valuable to future personal assistants after transaction costs), would we expect companies to collect this information from personal assistants and sell it? Maybe publishers and similar firms already do this. This could be a misleading example, any better ones?

People may forget a lot of information from college because most of the information they learn is useless for their job.

See also