The lesson is that to make best use of libraries as an education resource, their content has to be audited for reliability, and a system of classification introduced. Given that the library is already the main resource for students, the need is urgent. I suggest that an international consortium of universities should set up panels to audit the worth of books, endorsing those that are reliable. They should not censor, nor comment on matters of opinion--- the price we pay for the free press and free speech is the rubbish it produces. But they should authoritatively identify worthwhile books, and warn of factual error when it occurs. Without such expert monitoring, the library will increasingly be a problem rather than a boon, and limited in educational value.
OK, I'm lying. The quote is really about the Internet. Because nobody, you know, gets anything wrong in print media.
Leaving aside the desirability of such an endorsement strategy (I'm picturing little circa-1998 "web award for reliability" icons), what makes Grayling think his scheme is in any way practical? The example he gives is various incorrect transcriptions of a poem: does any one site get all poems (or lyrics) correct? The web is changeable: will a site that is "reliable" today be reliable tomorrow? How much effort does he think it would take to provide a bibliography of the web? Perhaps he isn't aware of previous efforts to provide human indexes?
The academic world would be better served by having more information freely available--- and that would be more in line with university goals than running truth commissions. References tied up behind paywalls won't be linked to, and certainly not used by casual researchers. Articles that don't appear online won't turn up in search engines. And anybody can create a bibliography (or improve Wikipedia's) if they want to improve the use of reliable sources.
Bonus fact-checking exercise: What is the atomic weight of Cesium?
132.90546: 22 pages on Google
132.90545: 403 pages
132.9054: 488 pages
132.9055: 77 pages
132.905: 714 pages
132.906: 8 pages, including a top-ranked hit from the CDC
How many pages on the minority side of each rounding decision look "unreliable"?
Extra credit homework: Go crack open Tufte, Visual Display of Quantitative Information, and look for the chart showing the thermal conductivity of copper vs. temperature. Or rather, all the published measurements, which vary by many orders of magnitude.