5 ms·

The trouble with Wikipedia is that it's an inch deep. For any given topic, especially history, there's a trove of information in scholarly books from good publi

by TillE 3y ago

The trouble with Wikipedia is that it's an inch deep. For any given topic, especially history, there's a trove of information in scholarly books from good publishers, but the corresponding Wikipedia article is like three pages long.

the_af 3y ago
Is it wrong that Wikipedia articles are only three pages long? Does anybody claim that reading an encyclopedia article (Wikipedia, Britannica or whatever) is better than reading a scholarly book on the given topic?
- Aerbil313 3y ago
  Some people just demand that Wikipedia be a universal factual info database, missing nothing. It'd be nice, though.
  - ghaff 3y ago
    There really isn't such a thing though for many topics as a universal factual info database. For many, e.g. historical, topics different books have different areas of focus and interpret events differently. Encyclopedias do to a certain degree (and historical "truth" may shift over time) but, in general, they're not the place to hash out the "right" interpretation of events.
    - LawTalkingGuy 3y ago
      > historical, topics different books have different areas of focus and interpret events differently In those areas the fact isn't the base fact, but claims of fact. We don't know who explored which bit of the great pyramid in which order, and may never, but we know of many specific claims. The fact check wouldn't be "The great pyramid X" but "Herodotus said X about the great pyramid". > in general, they're not the place to hash out the "right" interpretation of events. Once you scope the problem correctly it's not a problem. The point isn't to solve historical riddles, it's to document what evidence we have. Sometimes that evidence is broadly accepted measurements (land area of Australia) and other times it's not.
- NegativeK 3y ago
  People with just a Wikipedia-level knowledge will argue with actual experts as if Wikipedia is equivalent. Of course, you can't really say that distrusting experts is unique to encyclopedias.
  - the_af 3y ago
    But isn't that the same as people who argue because "they read it on a magazine/on the newspaper"? So they are wrong -- is it Wikipedia's fault though? An encyclopedia is always the starting point, never the end of serious research. (It's ok however to stick to Wikipedia if a superficial acquaintance with the topic is enough!).
  - harshreality 3y ago
    If there's something in a wikipedia article that experts will argue against, the article needs updating to be compatible with, even if does not include, expert-level knowledge.
mfer 3y ago
Wikipedia is a summary. Is it meant to be deep? If you want to go deep on any topic you'll need to go to other sources.
Wissenschafter 3y ago
Isn't that literally the point of an encyclopedia? A starting point, it's the abstract on the subject if you will.
- seydor 3y ago
  it has a good format though and would be nice to have a second level of scholarship (e.g scholarpedia). Modeling itself after encyclopedia would be regressive
- tivert 3y ago
  > Isn't that literally the point of an encyclopedia? A starting point, it's the abstract on the subject if you will. Yes, but Wikipedia is also frequently conceived and marketed as "the sum of all human knowledge," which that shows is a lie by definition.
  - jjoonathan 3y ago
    If we could condemn a thing due to hype alone, we would condemn all that is good in the world.
  - humanistbot 3y ago
    > "the sum of all human knowledge," which that shows is a lie by definition. By which definition? In math, the sum of a set necessarily implies a loss of information about that set, for sets larger than 1. But they're using "sum" not in the purely mathematical sense, more like "the summary of all human knowledge". But the same principle applies axiomatically, because summaries are lossy compression: you cannot have a summary that contains all the information of the source it is summarizing.
    - tivert 3y ago
      >> "the sum of all human knowledge," which that shows is a lie by definition. > By which definition? That's pretty easy: definition 2 "the whole amount : aggregate" (https://www.merriam-webster.com/dictionary/sum https://www.merriam-webster.com/dictionary/sum). That it's interpreted that way is shown by the frequency of people saying stuff like "I loaded Wikipedia onto this battery powered Raspberry Pi in a Pelican case, now I'm ready to rebuild civilization if it collapses," and seemingly believing it. But you do correctly point to another issue: sum has a meaning of "a summary of the chief points or thoughts," which I feel is a less common usage. So the marketing phrase may not be so much a lie, but rather an extremely misleading statement that invites misinterpretation that usually goes unchallenged. IMHO, those are actually even more pernicious than outright lies.
xiphias2 3y ago
Maybe a Wikipedia based LLM could make decisions about which papers are factual enough to include in a more extensive LLM.
- joeframbach 3y ago
  You'll end up with a corpus consisting entirely of paywall text and 404 pages.
yieldcrv 3y ago
Check the wiki page in another language, closer to the affected area. Its not a direct translation, its an entirely different encyclopedia and can be far more robust. (Maybe an LLM could harmonize all the wikipedias across languages)
jojobas 3y ago
Can there be factual correctness from being grounded in scholarly books as such? The amount of disagreement between researchers over time and changing consensus requires an external arbiter of individual facts at the very least.
bsaul 3y ago
Just an anecdotal experience : wikipedia is up-to-date with regards to "recent" discoveries. Whereas books will always be engraved with what was regarded as knowledge at the time of writing. Case in point : elder of zions protocol french article contained outdated knowledge (it said we knew who the author was, propagating an old hypothesis that had been debunked in the last 10 years, eventhough the other wikipedia articles were fixed). Historians were heard repeating that same boggus claim on radio. Until i convinced an historian friend of mine to fix the french article, and all of the sudden historians started fixing their speech. Meaning not only did they not update their knowledge from scholarly books, but they needed wikipedia to help them get up to date.
- quotemstr 3y ago
  New is not always better. For example, from about 1960-2010, anthropologists universally held a "pots, not people" view of prehistory: they asserted, with great confidence, that styles of pottery and metalworking changed over time due to voluntary exchange of ideas among peaceful, cooperating peoples. These anthropologists asserted that pre-1960 theories that pottery styles changed because population groups violently replaced each other were not only wrong, but immoral and barbaric. To them, it was modernity that made humans violent. Now due to ancient DNA, we know that the pre-1960s anthropologists were right and the post-1960 consensus was wrong: prehistory was violent and populations violently replaced each other with regularity. You're more informed reading, say, Gordon Child book from 1920 than a serious book on prehistoric archaeology from 2000. So it goes in many fields. Imagine how much longer it would take for science's self correction mechanism to operate if our knowledge were encoded solely in a "living" information system aligned with only currently fashionable ideas.
  - ragequitta 3y ago
    But wouldn't you agree reading about this topic now, with the counter-argument of the post-1960 consensus (though I have a hard time thinking most things debatable like this are ever strictly consensus), and the follow-up DNA evidence, is far more informative and convincing than what you would read in 1920? It seems that the people guessing from 1920 might've had about as much chance of being right as the people guessing in 1960 with neither having the relevant evidence to back their claim.
wing-_-nuts 3y ago
I don't think that's a fair criticism of wikipedia. Summarizing knowledge is literally the job of an encyclopedia. There was a reason all my professors in college told us to use wiki as a jumping off point for further reading in the citations.
- arp242 3y ago
  I think it's a fair criticism; or rather, an important limitation one needs to keep in mind. Wikipedia articles can miss out on a great deal of nuance and context, which can matter a great deal.
  - thebooktocome 3y ago
    Worse than that, motivated editors often color their pet pages with specific nuance and context, and no casual editor has a hope of winning an edit war against such opposition. My favorite example of this is the debate between a faction that believes Lithobates is the proper genus of a certain set of frogs, and another faction that believes the correct genus is Rana. The Lithobates side is essentially one person along with his sock and/or meat puppets, so in the end, after many rounds of moderation, most of the species in question are listed under both genuses.
  - jurimasa 3y ago
    [flagged]
- Closi 3y ago
  > I don't think that's a fair criticism of wikipedia. It's not a fair criticism of Wikipedia, but it is a fair criticism of using Wikipedia as a single-source.
- mcpackieh 3y ago
  Once you stray off the popsci/undergrad topics, wikipedia's summation of knowledge is often a few sentences if any. Topics for which numerous books have been written may get only one or two sentences on wikipedia, so I think it's fair to say that wikipedia is an inch deep. Maybe a few inches deep since popular topics do get longer articles, but the long tail of knowledge gets very shallow coverage on wikipedia.
curiousllama 3y ago
I wonder if that's the next iteration of Wikipedia. Right now, the model is to summarize secondary sources. Once summarization becomes trivial via LLMs, the most valuable thing to do would be to assemble ever-expanding datasets of secondary sources for the LLM to pull from.
weregiraffe 3y ago
Wikipedia is full of references to scholarly sources. Make a bot that follows the references to the sources and incorporates them in the training data, and Bob's your uncle.
klyrs 3y ago
My hope in this regard is that Wikipedia pages tend to have much more than an inch of citations. If even a significant fraction of those sources can be digested, it could give rise to a much deeper source. The really cool thing about their chatbot is that it appears to have the ability to summarize and highlight where the summaries came from. Extending that to the ability to summarize the backing sources, and point to where that came from, could be an incredible research tool.
ilyt 3y ago
Well, it has link that it can follow. Add some AI to take the footnotes and get to the sources and train on that.
mcv 3y ago
Still deeper than most people, though. You can't put all of human knowledge in Wikipedia, but it's extremely thorough in the basics.
narag 3y ago
Please, before assuming you know what I mean, read the complete comment. The trouble with Wikipedia is that it's an inch deep. That's not its biggest problem. Wikipedia is biased. Of course there's political bias in the American fashion. But that's not all. There is bias about History depending on what country is telling the story. And there is a strong bias even in scientific topics (maybe specially in them) when there are commercial interests involved. That's not specific to Wikipedia. But when you research some topic, reading multiple books, you'll notice there are different opinions, you learn to discount bias looking at the procedence. Wikipedia tries to adopt a neutral tone and cite different sources, but sometimes it does a terrible job at it.
- nomel 3y ago
  > There is bias about History depending on what country is telling the story. A funny example of this is fan death [1]. Comparing English to southeast Asian languages show that Asian languages pages suggest that it's real (at least last few times I translated it). An, in a bit of relevancy, the Japanese page has been overwritten with "I love you" and "I'm sorry". [1] https://en.wikipedia.org/wiki/Fan_death https://en.wikipedia.org/wiki/Fan_death
  - mr_toad 3y ago
    Well that was bizarre. But if you have a government agency saying that electric fans can cause asphyxiation, and even that they have data to provide it, can you blame people in that country for believing it?
janalsncm 3y ago
Wikipedia is not some thing handed down by God at the beginning of time. It’s a work in progress by volunteers. If you think a page lacks depth, you’re free to update it.
mr_toad 3y ago
If anything Wikipedia has a problem with not being able to summarise and having too much original research. Some articles are just pages and pages of formulas with few citations.

by TillE 3y ago

the_af 3y ago

Is it wrong that Wikipedia articles are only three pages long? Does anybody claim that reading an encyclopedia article (Wikipedia, Britannica or whatever) is better than reading a scholarly book on the given topic?

Aerbil313 3y ago
Some people just demand that Wikipedia be a universal factual info database, missing nothing. It'd be nice, though.
- ghaff 3y ago
  There really isn't such a thing though for many topics as a universal factual info database. For many, e.g. historical, topics different books have different areas of focus and interpret events differently. Encyclopedias do to a certain degree (and historical "truth" may shift over time) but, in general, they're not the place to hash out the "right" interpretation of events.
  - LawTalkingGuy 3y ago
    > historical, topics different books have different areas of focus and interpret events differently In those areas the fact isn't the base fact, but claims of fact. We don't know who explored which bit of the great pyramid in which order, and may never, but we know of many specific claims. The fact check wouldn't be "The great pyramid X" but "Herodotus said X about the great pyramid". > in general, they're not the place to hash out the "right" interpretation of events. Once you scope the problem correctly it's not a problem. The point isn't to solve historical riddles, it's to document what evidence we have. Sometimes that evidence is broadly accepted measurements (land area of Australia) and other times it's not.
NegativeK 3y ago
People with just a Wikipedia-level knowledge will argue with actual experts as if Wikipedia is equivalent. Of course, you can't really say that distrusting experts is unique to encyclopedias.
- the_af 3y ago
  But isn't that the same as people who argue because "they read it on a magazine/on the newspaper"? So they are wrong -- is it Wikipedia's fault though? An encyclopedia is always the starting point, never the end of serious research. (It's ok however to stick to Wikipedia if a superficial acquaintance with the topic is enough!).
- harshreality 3y ago
  If there's something in a wikipedia article that experts will argue against, the article needs updating to be compatible with, even if does not include, expert-level knowledge.

mfer 3y ago

Wikipedia is a summary. Is it meant to be deep? If you want to go deep on any topic you'll need to go to other sources.

Wissenschafter 3y ago

Isn't that literally the point of an encyclopedia? A starting point, it's the abstract on the subject if you will.

seydor 3y ago
it has a good format though and would be nice to have a second level of scholarship (e.g scholarpedia). Modeling itself after encyclopedia would be regressive
tivert 3y ago
> Isn't that literally the point of an encyclopedia? A starting point, it's the abstract on the subject if you will. Yes, but Wikipedia is also frequently conceived and marketed as "the sum of all human knowledge," which that shows is a lie by definition.
- jjoonathan 3y ago
  If we could condemn a thing due to hype alone, we would condemn all that is good in the world.
- humanistbot 3y ago
  > "the sum of all human knowledge," which that shows is a lie by definition. By which definition? In math, the sum of a set necessarily implies a loss of information about that set, for sets larger than 1. But they're using "sum" not in the purely mathematical sense, more like "the summary of all human knowledge". But the same principle applies axiomatically, because summaries are lossy compression: you cannot have a summary that contains all the information of the source it is summarizing.
  - tivert 3y ago
    >> "the sum of all human knowledge," which that shows is a lie by definition. > By which definition? That's pretty easy: definition 2 "the whole amount : aggregate" (https://www.merriam-webster.com/dictionary/sum https://www.merriam-webster.com/dictionary/sum). That it's interpreted that way is shown by the frequency of people saying stuff like "I loaded Wikipedia onto this battery powered Raspberry Pi in a Pelican case, now I'm ready to rebuild civilization if it collapses," and seemingly believing it. But you do correctly point to another issue: sum has a meaning of "a summary of the chief points or thoughts," which I feel is a less common usage. So the marketing phrase may not be so much a lie, but rather an extremely misleading statement that invites misinterpretation that usually goes unchallenged. IMHO, those are actually even more pernicious than outright lies.

xiphias2 3y ago

Maybe a Wikipedia based LLM could make decisions about which papers are factual enough to include in a more extensive LLM.

joeframbach 3y ago
You'll end up with a corpus consisting entirely of paywall text and 404 pages.

yieldcrv 3y ago

Check the wiki page in another language, closer to the affected area. Its not a direct translation, its an entirely different encyclopedia and can be far more robust. (Maybe an LLM could harmonize all the wikipedias across languages)

jojobas 3y ago

Can there be factual correctness from being grounded in scholarly books as such? The amount of disagreement between researchers over time and changing consensus requires an external arbiter of individual facts at the very least.

bsaul 3y ago

Just an anecdotal experience : wikipedia is up-to-date with regards to "recent" discoveries. Whereas books will always be engraved with what was regarded as knowledge at the time of writing. Case in point : elder of zions protocol french article contained outdated knowledge (it said we knew who the author was, propagating an old hypothesis that had been debunked in the last 10 years, eventhough the other wikipedia articles were fixed). Historians were heard repeating that same boggus claim on radio. Until i convinced an historian friend of mine to fix the french article, and all of the sudden historians started fixing their speech. Meaning not only did they not update their knowledge from scholarly books, but they needed wikipedia to help them get up to date.

quotemstr 3y ago
New is not always better. For example, from about 1960-2010, anthropologists universally held a "pots, not people" view of prehistory: they asserted, with great confidence, that styles of pottery and metalworking changed over time due to voluntary exchange of ideas among peaceful, cooperating peoples. These anthropologists asserted that pre-1960 theories that pottery styles changed because population groups violently replaced each other were not only wrong, but immoral and barbaric. To them, it was modernity that made humans violent. Now due to ancient DNA, we know that the pre-1960s anthropologists were right and the post-1960 consensus was wrong: prehistory was violent and populations violently replaced each other with regularity. You're more informed reading, say, Gordon Child book from 1920 than a serious book on prehistoric archaeology from 2000. So it goes in many fields. Imagine how much longer it would take for science's self correction mechanism to operate if our knowledge were encoded solely in a "living" information system aligned with only currently fashionable ideas.
- ragequitta 3y ago
  But wouldn't you agree reading about this topic now, with the counter-argument of the post-1960 consensus (though I have a hard time thinking most things debatable like this are ever strictly consensus), and the follow-up DNA evidence, is far more informative and convincing than what you would read in 1920? It seems that the people guessing from 1920 might've had about as much chance of being right as the people guessing in 1960 with neither having the relevant evidence to back their claim.

wing-_-nuts 3y ago

I don't think that's a fair criticism of wikipedia. Summarizing knowledge is literally the job of an encyclopedia. There was a reason all my professors in college told us to use wiki as a jumping off point for further reading in the citations.

arp242 3y ago
I think it's a fair criticism; or rather, an important limitation one needs to keep in mind. Wikipedia articles can miss out on a great deal of nuance and context, which can matter a great deal.
- thebooktocome 3y ago
  Worse than that, motivated editors often color their pet pages with specific nuance and context, and no casual editor has a hope of winning an edit war against such opposition. My favorite example of this is the debate between a faction that believes Lithobates is the proper genus of a certain set of frogs, and another faction that believes the correct genus is Rana. The Lithobates side is essentially one person along with his sock and/or meat puppets, so in the end, after many rounds of moderation, most of the species in question are listed under both genuses.
- jurimasa 3y ago
  [flagged]
Closi 3y ago
> I don't think that's a fair criticism of wikipedia. It's not a fair criticism of Wikipedia, but it is a fair criticism of using Wikipedia as a single-source.
mcpackieh 3y ago
Once you stray off the popsci/undergrad topics, wikipedia's summation of knowledge is often a few sentences if any. Topics for which numerous books have been written may get only one or two sentences on wikipedia, so I think it's fair to say that wikipedia is an inch deep. Maybe a few inches deep since popular topics do get longer articles, but the long tail of knowledge gets very shallow coverage on wikipedia.

curiousllama 3y ago

I wonder if that's the next iteration of Wikipedia. Right now, the model is to summarize secondary sources. Once summarization becomes trivial via LLMs, the most valuable thing to do would be to assemble ever-expanding datasets of secondary sources for the LLM to pull from.

weregiraffe 3y ago

Wikipedia is full of references to scholarly sources. Make a bot that follows the references to the sources and incorporates them in the training data, and Bob's your uncle.

klyrs 3y ago

My hope in this regard is that Wikipedia pages tend to have much more than an inch of citations. If even a significant fraction of those sources can be digested, it could give rise to a much deeper source. The really cool thing about their chatbot is that it appears to have the ability to summarize and highlight where the summaries came from. Extending that to the ability to summarize the backing sources, and point to where that came from, could be an incredible research tool.

ilyt 3y ago

Well, it has link that it can follow. Add some AI to take the footnotes and get to the sources and train on that.

mcv 3y ago

Still deeper than most people, though. You can't put all of human knowledge in Wikipedia, but it's extremely thorough in the basics.

narag 3y ago

Please, before assuming you know what I mean, read the complete comment. The trouble with Wikipedia is that it's an inch deep. That's not its biggest problem. Wikipedia is biased. Of course there's political bias in the American fashion. But that's not all. There is bias about History depending on what country is telling the story. And there is a strong bias even in scientific topics (maybe specially in them) when there are commercial interests involved. That's not specific to Wikipedia. But when you research some topic, reading multiple books, you'll notice there are different opinions, you learn to discount bias looking at the procedence. Wikipedia tries to adopt a neutral tone and cite different sources, but sometimes it does a terrible job at it.

nomel 3y ago
> There is bias about History depending on what country is telling the story. A funny example of this is fan death [1]. Comparing English to southeast Asian languages show that Asian languages pages suggest that it's real (at least last few times I translated it). An, in a bit of relevancy, the Japanese page has been overwritten with "I love you" and "I'm sorry". [1] https://en.wikipedia.org/wiki/Fan_death https://en.wikipedia.org/wiki/Fan_death
- mr_toad 3y ago
  Well that was bizarre. But if you have a government agency saying that electric fans can cause asphyxiation, and even that they have data to provide it, can you blame people in that country for believing it?

janalsncm 3y ago

Wikipedia is not some thing handed down by God at the beginning of time. It’s a work in progress by volunteers. If you think a page lacks depth, you’re free to update it.

mr_toad 3y ago

If anything Wikipedia has a problem with not being able to summarise and having too much original research. Some articles are just pages and pages of formulas with few citations.