Golden: Mapping human knowledge • Hacker News Board

Re: Golden: Mapping human knowledge

#101

Post by judegomila » Tue, Apr 30, 2019, 10:26 PM UTC

msla wrote: post #86

llamataboot wrote: post #73

Cynical self: They raised a $5mil round just to re-do wikipedia with more self-referential VC pats on the back. Intrigued self: If there is anything worth raising money for, it is building out a collective knowledge base for all of humanity. Angry self: Wikipedia is quite literally one of the most amazing things that the internet and millions of strangers have ever produced together, and lives as a vital proof of con…

> Also, what are your thoughts on the morality of essentially crowdsourcing all the input that you will feed to your query engines and profit off people using? Aren't you essentially making your contributors all unpaid workers? My mind went in a different direction, although your point is very valuable and the question needs to be asked. My concern is this: Wikipedia editors work hard to maintain a level of quality h…

Jude from Golden here. Yes we will work very hard to keep to scientific consensus and reach that level of quality or higher. Please bear in mind this is not only crowdsourcing but there is automation in the collection of the information, that too will have its fair share of issues as well no doubt :> There is likely opportunity to auto detect cranky/racist/gaming behavior and we have some plans in that area. Feel free to shoot over ideas as well if you have further thoughts.

Re: Golden: Mapping human knowledge

#102

Post by yorwba » Tue, Apr 30, 2019, 10:28 PM UTC

jake-low wrote: post #95

hk__2 wrote: post #83

Earlier quoted context omitted.

Given than Wikidata is published under a CC 0 licence there’s nothing preventing Golden from using it as a knowledge source. Unless Golden provides a better editor experience I see no reason for anyone to contribute to Golden rather than Wikidata.

Just FYI, most of the content on Wikipedia is actually dual licensed under CC-BY-SA and GFDL. Both licenses are copyleft licenses; CC0 is more permissive. https://en.wikipedia.org/wiki/Wikipedia:Reusing_Wikipedia_co...

Parts of Wikipedia articles (e.g. some infoboxes [1]) are actually sourced from Wikidata, which is licensed under CC0 [2]

[1] https://en.wikipedia.org/wiki/Template:Infobox_person/Wikida...

[2] https://www.wikidata.org/wiki/Wikidata:Data_access#Basic_imp...

Re: Golden: Mapping human knowledge

#103

Post by llamataboot » Tue, Apr 30, 2019, 10:33 PM UTC

judegomila wrote: post #98

llamataboot wrote: post #73

Cynical self: They raised a $5mil round just to re-do wikipedia with more self-referential VC pats on the back. Intrigued self: If there is anything worth raising money for, it is building out a collective knowledge base for all of humanity. Angry self: Wikipedia is quite literally one of the most amazing things that the internet and millions of strangers have ever produced together, and lives as a vital proof of con…

Jude from Golden here. Agreed that WP is one of the most amazing things ever built and interesting to see your various lenses on our mission. To the cynical self: see dropbox launch on HN back in the day. PS I’m no way claiming we are dropbox :> To the angry self: There are various constraints that we want to release ourselves from in working on this problem by starting fresh. We believe the constraint space is too h…

Thanks for replying! I agree micropayments are incredibly difficult to make work, as is evidenced by the race to the bottom advertising model that seems to be everywhere all-the-time.

I hear from you that you want to essentially cover 10 billion topics, and essentially validate that they exist, but that says nothing about validating the content of what someone is saying about it, nor organizing it, etc.

I hear lots of AI buzzwords, but essentially I don't see any staff that would leverage all the thinking humanity has done on organizing, validating, and cataloging information. Where are the information scientists? The librarians? The archivists? The journalists? The philosophers? Etc etc etc.

Essentially you are talking about a profoundly /human/ endeavor that requires input (IMHO) from many corners of human knowledge to do in any way that begins to approach wikipedia (or even an encyclopedia, much less a library) in terms of quality and scale.

I hear buzzwords, and see an alarming lack of acknowledgement of how difficult these questions are (or even that they exist).

However, you have the $$ and the people, and I'm just here hiding behind a keyboard criticizing. Clearly you've convinced more people of your ideas than I have of mine, and by all means it's a noble goal, so I wish you the best of luck and will be interesting to see what you are and what Golden looks like in a decade or so!

Re: Golden: Mapping human knowledge

#104

Post by judegomila » Tue, Apr 30, 2019, 10:33 PM UTC

pradn wrote: post #69

judegomila wrote: post #55

Earlier quoted context omitted.

Jude from Golden here. This knowledge base is for all of humanity as well. I think the important part here is that we are putting out the content on CC-BY-SA-4.0. There is also risk factor to a donations model for WP and to not having enough revenue to invest in the tools needed to go 1000x on the topics + other features that are important to build. If it means anything, I worked on my last co (Heyzap) for 9 years an…

Thank you for your response. I'm greatly encouraged that the license is CC-BY-SA. That's definitely in the public good. I do believe there is space for non-Wikipedia knowledge repositories. Look at Wikia, for example. People do want to store all their fandom knowledge somewhere. It's just that Wikipedia might not be the place for it. I also agree that the underinvestment in tools can limit contributions. Wikipedia's…

Thanks, glad you like it.

Re: Golden: Mapping human knowledge

#105

Post by llamataboot » Tue, Apr 30, 2019, 10:34 PM UTC

judegomila wrote: post #101

msla wrote: post #86

Earlier quoted context omitted.

> Also, what are your thoughts on the morality of essentially crowdsourcing all the input that you will feed to your query engines and profit off people using? Aren't you essentially making your contributors all unpaid workers? My mind went in a different direction, although your point is very valuable and the question needs to be asked. My concern is this: Wikipedia editors work hard to maintain a level of quality h…

Jude from Golden here. Yes we will work very hard to keep to scientific consensus and reach that level of quality or higher. Please bear in mind this is not only crowdsourcing but there is automation in the collection of the information, that too will have its fair share of issues as well no doubt :> There is likely opportunity to auto detect cranky/racist/gaming behavior and we have some plans in that area. Feel fre…

"there is likely opportunity to auto detect cranky/racist/gaming behavior and we have some plans in that area."

Ask facebook and twitter how easy automated content moderation is...

Re: Golden: Mapping human knowledge

#106

Post by hk__2 » Tue, Apr 30, 2019, 10:49 PM UTC

jake-low wrote: post #95

hk__2 wrote: post #83

Earlier quoted context omitted.

Given than Wikidata is published under a CC 0 licence there’s nothing preventing Golden from using it as a knowledge source. Unless Golden provides a better editor experience I see no reason for anyone to contribute to Golden rather than Wikidata.

Just FYI, most of the content on Wikipedia is actually dual licensed under CC-BY-SA and GFDL. Both licenses are copyleft licenses; CC0 is more permissive. https://en.wikipedia.org/wiki/Wikipedia:Reusing_Wikipedia_co...

Yes, but I’m talking about Wikidata (CC0), not Wikipedia (CC-BY-SA).

Re: Golden: Mapping human knowledge

#107

Post by judegomila » Tue, Apr 30, 2019, 10:49 PM UTC

llamataboot wrote: post #103

judegomila wrote: post #98

Earlier quoted context omitted.

Jude from Golden here. Agreed that WP is one of the most amazing things ever built and interesting to see your various lenses on our mission. To the cynical self: see dropbox launch on HN back in the day. PS I’m no way claiming we are dropbox :> To the angry self: There are various constraints that we want to release ourselves from in working on this problem by starting fresh. We believe the constraint space is too h…

Thanks for replying! I agree micropayments are incredibly difficult to make work, as is evidenced by the race to the bottom advertising model that seems to be everywhere all-the-time. I hear from you that you want to essentially cover 10 billion topics, and essentially validate that they exist, but that says nothing about validating the content of what someone is saying about it, nor organizing it, etc. I hear lots o…

Thanks so much, in terms of the hard questions after todays madness of launch comes down a little I'll tackle the hardest comments/questions here. In the coming months we will do some technical blog posts to explain how we will tackle the problem space. Many of the problems we have not figured out yet and welcome the community to contact us with new ideas. I 100% agree some of the problems are very hard. In terms of giving a glimpse into some problems we have solved so far, please test out the AI assisted editor, the magic table cells in the editor for auto filling tables, the citation tool by pasting a academic paper in the citation UI, the event detection on the timeline UI and the AI suggestions as well to get to some of the early results we have on automating the problem. Topic prediction, taxonomic detection, claim validation, structured data extraction, auto field detection/suggestions, crosslinking, spelling/grammar checking, sentiment checking, event detection, tense detection, quality on human edit feedbacks and ultimately prose writing (see recent open AI auto writing research) [non exhaustive list] - some we have solved and some not yet, but we will keep working on it. Generally speaking, keen to work on something difficult for the next 10 years...

Re: Golden: Mapping human knowledge

#108

Post by timClicks » Tue, Apr 30, 2019, 10:56 PM UTC

judegomila wrote: post #98

llamataboot wrote: post #73

Cynical self: They raised a $5mil round just to re-do wikipedia with more self-referential VC pats on the back. Intrigued self: If there is anything worth raising money for, it is building out a collective knowledge base for all of humanity. Angry self: Wikipedia is quite literally one of the most amazing things that the internet and millions of strangers have ever produced together, and lives as a vital proof of con…

Jude from Golden here. Agreed that WP is one of the most amazing things ever built and interesting to see your various lenses on our mission. To the cynical self: see dropbox launch on HN back in the day. PS I’m no way claiming we are dropbox :> To the angry self: There are various constraints that we want to release ourselves from in working on this problem by starting fresh. We believe the constraint space is too h…

So, you're aiming for an exit as per Metaweb/Freebase?

Re: Golden: Mapping human knowledge

#109

Post by judegomila » Tue, Apr 30, 2019, 10:56 PM UTC

llamataboot wrote: post #105

judegomila wrote: post #101

Earlier quoted context omitted.

Jude from Golden here. Yes we will work very hard to keep to scientific consensus and reach that level of quality or higher. Please bear in mind this is not only crowdsourcing but there is automation in the collection of the information, that too will have its fair share of issues as well no doubt :> There is likely opportunity to auto detect cranky/racist/gaming behavior and we have some plans in that area. Feel fre…

"there is likely opportunity to auto detect cranky/racist/gaming behavior and we have some plans in that area." Ask facebook and twitter how easy automated content moderation is...

True, very hard problem. But combinatorial opinion free form comments phase space >> than phase space of canonical knowledge thus number of patterns of things that can go wrong for them much larger. Still keeping our shields up and not discounting this issue.

Re: Golden: Mapping human knowledge

#110

Post by teddyh » Tue, Apr 30, 2019, 10:57 PM UTC

ryan_lane wrote: post #77

teddyh wrote: post #27

> Today, we have a great opportunity to use new technologies to solve the problems: […] real identity […] > Join Golden > First Name Last Name Yeah, no. A “real names” policy is a really terrible idea , for reasons outlined here: https://geekfeminism.wikia.org/wiki/Who_is_harmed_by_a_%22Re...

(former wikimedia employee here) A real names policy also means it'll have essentially no Japanese content. Online psudonyms, or anonymous editing are the norm in Japan.

That’s listed in the linked text:

> in some countries, such as Japan, online pseudonyms are the norm in all circumstances