Knowledge Aggregation & Propagation
Includes merged notes from Rhys
I'll be the contrarian - I think academia is awful and should be destroyed.
Reputation is completely independent of content and should not be the mechanism by which you judge the quality of research; your reputation is worthless to me. (laughter)
My perspective1: a Bitcoin developer perspective. I work with Bitcoin Core developers and know this side of the ecosystem. worked at LedgerX (4 years). I also work on biotech on the hobbyist side, garage engineering -- what I was doing when bitcoin was launched. This is a community that recognized there are amazing advances in science, you can do all sorts of things on your own. You don't need to be on the manuscript treadmill and go to a university lab to do this science.
Basically, this is an overview of resources from the Bitcoin community. Mostly I’ll talk about transcripts. And continue to talk about how terrible academia is.
All conferences should have them! This is a great way to respect your audience. The logistics of going to conference talks and conferences is itself ridiculous. The economic cost of going to confs is quite high. You should multiply the # of participants by $1000 to figure out how much they spent — more if you include hotels and food.
Text is more convenient for mass consumption, takes up less of people’s time. Do it real-time and publish immediately after. I started doing this way back in high school to try prove the content being delivered in high school was awful and not worth my time. I did this for every class for 4 years. It turns out no one cares. Kind of of how I felt about high school itself.
One thing I’ve noticed is how important timing is. I publish the transcript by the time the speaker sits down. I’ve experimented with this: If I wait a day or two about transcripts, no one cares. They like having them immediately.
I’ve done a bunch of transcripts.
There’s arcane knowledge at conferences that isn’t communicated elsewhere. Not everyone can attend. Or for people who don’t know to which talks to attend in the 1st place, how would they even know [before they are transcribed + searchable]? I’ve done Scaling Bitcoin, MIT Bitcoin Exp for 5 years, including Bitcoin Edge Dev++ for training developers in basic and advanced structures — also available as videos.
People seem to like transcripts; it is very competitive in a funny way. They take videos and post them. It’s fun to do, I don’t just do it for the knowledge. (hi Jim Carrey!) Here’s an image [of transcription in progress] that’s recursive. It’s art!
Tools and uses
I use markdown, git, and Ikiwiki. I also do bookmarking from the commandline: using jotmuch/buku. It’s hard to rememer urls, so tag bookmarks and do a tag search.
[How are transcripts used?] Here’s a google scholar search [for my wiki], excluding the site hosting the transcript, and find other links to them. Human genome discussions, FBI interest, then some nice bitcoin papers referencing the transcripts in formal publication, and patents!
Over 10y of doing this, I have 1.5M words and 10MB of text, around 650 talk transcripts. If you have questions like “have any bitcoin devs considered how the fee market would work without a subsidy?” the answer is “yes”.
How can we get more scribes?
If I’m going to be at a conference anyway, I might as well type transcripts. But I can’t be everywhere. Other people should do this; we should help fund people other than me to go to confs and type things. Commercial / court stenographers make a lot of money - similar to software developers (easily 200k/y).
Open research questions
What about machine learning for speech recognition? I tried this, first approach had a 20% error rate w/ word recognition. Mozilla later redid the same algorithm and made a better implementation; but both are trained on audiobooks, not conference speakers.
Improving conference quality: Use surveys before/after. Has anyone studied this? I’ve been to only 2 confs that have done this.
Conference submissions tools need improvement. With program committees I have been on, Easychair is commonly used [but not ideal]. We should select talks more carefully. Commission or invite specific talks from experts — don’t just hope that people who hear about and can attend the conf submit something relevant. Small improvements here have large effects.
Bridges/resources for academia
bitcoin-dev mailing list, IRC logs (bitcoin-wizards, bitcoin-core-dev), and bitcointalk.org
Hosted by the Linux Foundation. It turns out they don’t like email anymore and are deprecating supporting email lists, so we’re looking for another solution.
If you’re thinking about studying something, find out who has done it before: and read their work /go talk to them [I encourage you do to both] Search on the above sites, it often won’t be found on Google Scholar.
Asides and alternatives
Proof of Stake?
I see there’s a lot of talk about PoS here [at this conference]. We looked into that years ago and proved that it doesn’t work. It’s like reputation; it isn’t costly and doesn’t work. [people should read about past discussions]
Back up your data! 8 yrs ago I lost all my marks in a tragic boating accident.
Alternatives to transcripts
(Speculative) Meetlogs— a homebrew crm, record all conversations, track who I talked to and about what, trace the origin of ideas, and their frequency and contagion. Like a rolodex.
This is especially powerful in the Bitcoin communities: we have high bus-factors, if those people leave the knowledge goes away. Track the sharding factor for ideas people care about.
Alternatives to academia
Academia is broken. Publishing in particular. What to do?
Today we talked about reputation tokens; prediction markets, derivatives markets… For access to knowledge: Sci-hub is amazing. But it’s illegal today. [thanks to broken publishing intermediaries]
~ I don’t understand how prediction markets could help here, but [Robin Hanson] often talks about this [as an alternative to publisher-mediated reputation from closed publications]. Someone has to fund your prediction; isn’t the fundamental problem getting funding in the first place? And you need reputation again to inspire prediction…
Sci-hub: I may have accidentally doxxed Alexandra Elbakyan — recognized her profile picture because I had seen her at a Harvard conf in 2010. I said that in public…. I really like sci-hub! but it does have problems. Namely: I’ve been looking for a complete copy of the 50TB dataset. Sci-hub is the only entity that has it, apparently. Librarians should have copies but they all seem to have sold out to OCLC. I also have a rant against libraries, obviously!
Increase connection from academics to bitcoin devs
I encourage people working on blockchain projects to reach out to Bitcoin core devs in particular. Old concepts posted on forums and lists only sometimes get cited; is often folklore but still a valuable source of knowledge.