Latest / Bitcoin And | Bitcoin & Economic News / Navigating the Social Graph
Transcript
- 0:00Unknown: Good morning. This is David Bennett, and this is Bitcoin and, a podcast where I try to find the edge effect between the worlds of Bitcoin, gaming, permaculture,
- 0:08podcasting, and education to gain a better understanding of all. Edge effect is a concept from ecology describing a greater diversity of life where the edges of 2 systems overlap. While species from either system can be found at the edge, it is important to note there are species in the overlap that exist in neither system, and that is what I seek to uncover. Uncover. So join me in discovering the variety of things being created as Bitcoin rubs up against other systems.
- 0:44It is 9:35
- 0:45AM Pacific Daylight Time. It is the 18th day of June 2024, and this is episode
- 0:51909
- 0:53of Bitcoin. And today, I'm gonna do something completely different.
- 0:58I'm going to bring you a reading of
- 1:01Papalia's
- 1:04Navigating
- 1:05the Social Graph. Now, I found this article,
- 1:08yesterday or the day before, and I kinda found it fascinating because it talks about things like web of trust and identity
- 1:15and denial of service attacks
- 1:18on your identity
- 1:19as somebody who's using Nostr, but it, you know, could be Facebook, it could be Twitter.
- 1:24This is like sort of the the social
- 1:26graph problem.
- 1:29And this piece goes into
- 1:31what is a social graph?
- 1:33And it's it's very good. However, there are there are some problems with it, not in its ideas, but in the fact that it is
- 1:40it's heavily weighted
- 1:42towards
- 1:43matricy math. There's some algebra.
- 1:46There's using
- 1:48formulas to define certain terms,
- 1:50and I
- 1:52have done my best to be able to clip that out
- 1:56and not basically be sitting here saying
- 1:59a plus b equals, you know, c squared to the raised to the f sharp and all that stuff. So I've
- 2:06done my best to get that out of there with
- 2:10without
- 2:12destroying the central idea or the theme of this particular paper.
- 2:16Again,
- 2:17this paper is really good. And even if you don't know
- 2:21math, you should probably go read this paper. I will definitely be having
- 2:25the link to this paper in the show notes, and I highly recommend that you guys go read this because as we sail this
- 2:34novel river that is Nostra that we find ourselves on.
- 2:39Who we follow
- 2:41and what
- 2:42types of recommendations we make it make for each other.
- 2:46And that includes, like, who should you follow? Should you follow somebody just because I'm following them? Or should you follow somebody until, like, wait to follow somebody until I tell
- 2:57you, hey, you probably should also be following this person. Is there inferences that can be drawn?
- 3:03If I follow somebody
- 3:05and you follow somebody, but we're not following each other,
- 3:08do we have something in common?
- 3:11Well, this paper is is bringing that all home. But one of the most interesting parts about this paper is when we start talking about the w o t or web of trust
- 3:24and proof of work whether it's proof of work that I actually wrote the note or proof of work as to whether or not I actually generated my own keys.
- 3:33Like, because, like, right now
- 3:36I'm being spammed by followers
- 3:39and
- 3:40it it is the exact whoever wrote this bot for Noster,
- 3:44it's it's definitely the same person. Alright? It this is
- 3:49every day,
- 3:51I will get on to Noster,
- 3:52and at least 5 times during the day that I get on, I will notice that I have 27 new followers.
- 3:59But each one of these followers has almost the exact same
- 4:04avatar.
- 4:05It's like concentric orange circles
- 4:08that,
- 4:10more or less
- 4:11don't,
- 4:12you
- 4:13know, kinda don't line up, and they don't line up in certain patterns. It's sort of like think of, you know, the same clock face, but different pictures of the clock at different times.
- 4:24It's following only 17 people every time it follows somebody.
- 4:28So every time one of these things one of the whenever this thing gets activated,
- 4:34it spins up a new
- 4:36insect and n pub key pair
- 4:39and it follows the exact same 17 people.
- 4:42So now I've got like, I don't know, it started happening when I was right
- 4:46around 8,000 followers,
- 4:48and now I'm up to almost 12,000 followers, and it's not fun because they don't mean anything to me. I don't I don't care about these things. They they they have all of them have 0 posts, of course.
- 5:00They only follow these certain people.
- 5:03They never actually do or engage with Noster or the general population of Noster in any way
- 5:11other than inflating
- 5:13our follower counts.
- 5:15But what does that do to my web of trust?
- 5:17If something that nobody else
- 5:20constantly follow if if my web of trust is being diluted
- 5:24by a bunch of
- 5:26bots that don't make any sense or don't engage in any meaningful way.
- 5:31What does that what does that say about my account?
- 5:35Well, this paper goes through a lot of that stuff
- 5:38And I think we should probably be very, very attentive
- 5:42to how we start moving forward in the future because
- 5:45all of our trust is basically mitigated or provided by
- 5:50central parties,
- 5:523rd parties, to be exact, people like Elon Musk, people like Mark Zuckerberg, all these people. Right? They are providing some kind of
- 6:00nexus of which here's how you here's the way that you're going to interact with each other and here's the way you're going to discover each other and here's the way you're going to trust each other and I don't want any of that.
- 6:12However when you're in a completely free and open environment that Nostra provides
- 6:17it can be a problem.
- 6:19We are experiencing
- 6:20different problems.
- 6:22And
- 6:23as we move
- 6:25through Noster and on into the future,
- 6:28how are we going
- 6:30to fix our little red wagon? Well,
- 6:33Papillia,
- 6:34if that is the way that I can pronounce that name, he's also known as the social graph guy. He has Noster, so I will be putting his in pub
- 6:43into the show
- 6:44notes hopefully he's gonna kind of clear some of this up or at least give us some things
- 6:48to think about so with no further ado
- 6:52here it is.
- 6:53Navigating the Social Graph
- 6:56by Popela,
- 6:57the social graph guy.
- 7:00The emergence of decentralized
- 7:02social networks with sovereign identities,
- 7:04for example, Nostra,
- 7:06is a pivotal moment for the web,
- 7:09and it comes at the perfect time.
- 7:11In the age of artificial intelligence,
- 7:14where it's getting cheaper and cheaper to impersonate people,
- 7:18distort videos and images, or just make them up,
- 7:22how is it possible to know what's true?
- 7:24What does truth even mean these days?
- 7:28These are questions I will not attempt to answer. However,
- 7:32even if the ontology of truth has eluded philosophers forever,
- 7:36a more practical approach is needed if the web is not to degenerate
- 7:41into a nihilistic place where nothing is real, or worse,
- 7:46a place where everything requires identity verification
- 7:50and the approval
- 7:51of a central authority.
- 7:54In this paper,
- 7:55you will find a definition of the social graph,
- 7:59principles for thinking about it,
- 8:01and practical ideas for using it for DOS
- 8:04prevention,
- 8:05social discovery,
- 8:07anti impersonation,
- 8:09accurate ratings, and more.
- 8:11You will also learn about alternative mechanisms
- 8:14that can be used in conjunction with social graph analysis
- 8:18to provide more compelling
- 8:21and complete solutions.
- 8:24The social graph.
- 8:26First,
- 8:27what is a graph?
- 8:29A graph is a mathematical structure
- 8:32consisting of a set of nodes or vertices
- 8:35and a set of edges
- 8:37edges
- 8:38are defined by a relation
- 8:41on a set
- 8:43For example, if node v
- 8:46and node u
- 8:48are related,
- 8:49then the pair
- 8:51is contained in a set.
- 8:54The relationship can be symmetrical
- 8:57or asymmetrical,
- 8:58distinguishing between undirected and directed graphs.
- 9:01A social graph is simply a graph in which the nodes are human entities such as individuals,
- 9:07or companies, or groups.
- 9:10And the edges
- 9:11represent some form of social relationship
- 9:14such as being friends with, following, or being zapped.
- 9:19Before we go any further, we need to define another important notion.
- 9:23That
- 9:24of a neighborhood.
- 9:26A neighborhood is a set of vertices
- 9:29inside
- 9:30another set of vertices
- 9:32whose defining property
- 9:34is to be connected
- 9:35to a vertex
- 9:37in a set.
- 9:38What problem
- 9:40is the social graph solving?
- 9:42The answer is many, in many different ways and in many different forms.
- 9:46We are talking about using it as a defense mechanism against impersonation,
- 9:52as a way to prevent DOS attacks,
- 9:55to facilitate social discovery and connections, and to calculate
- 9:59more accurate ratings.
- 10:01But before we dive into the implications,
- 10:04I ask you to pause for a moment to understand the magnitude
- 10:07of what we are discussing
- 10:10Principles.
- 10:12Don't fall into the trap of thinking that the engineers at YouTube were evil when they designed their algorithm. As the saying goes, the road to hell is paved with good intentions.
- 10:24If your implementation is successful,
- 10:27people's behavior will be affected.
- 10:30That's why it's important to establish principles that can guide designers and developers in creating systems
- 10:36that are healthier
- 10:37than the world we're leaving behind.
- 10:41The principle
- 10:42of least interference.
- 10:44A good rule of thumb when designing systems that affect social dynamics is to follow the principle
- 10:49of least interference.
- 10:52The principle of least interference
- 10:54is this,
- 10:55solve the problem
- 10:57by implementing whatever mechanism has the least possible impact on social dynamics
- 11:03as a heuristic based on this principle it is generally better
- 11:07to focus on solving only one problem at a time since it is unlikely that solving multiple problems at once will not lead to a greater impact
- 11:17on those social dynamics.
- 11:19A great example to analyze is the
- 11:22WOT score implemented by Coracle,
- 11:25which tries to reduce spam, imposters,
- 11:29and objectionable content all at once.
- 11:33Let's start by analyzing how it works
- 11:36as an anti impersonation mechanism.
- 11:38If you search for Lynn Alden,
- 11:40it is very likely,
- 11:42if you are part of a, quote, good network, that the in pub with the highest
- 11:47WOT
- 11:48is the real
- 11:49LIN,
- 11:50while all others are impostors.
- 11:54However, a simple warning icon above all other LINS would have accomplished the same goal without introducing a rating that can alter social dynamics. In fact, the WOT score represents the opinion
- 12:07of the crowd around you, which doesn't offer any nuance when comparing the reputation
- 12:12of different real people.
- 12:15Crowd thinking is one of the biggest problems of the world that we are leaving, and it would be a silly exercise
- 12:21to bring it with us. Besides,
- 12:23we do not walk around with a number on our forehead in meat space,
- 12:28nor should we in cyberspace,
- 12:30even if that number depends on the observer.
- 12:34There is also another, more fundamental problem, which is that
- 12:38the user only has information
- 12:41about a limited subset of the network,
- 12:44as defined by the relation
- 12:46follow or mute.
- 12:48Therefore,
- 12:49if you set the WOT threshold above 0,
- 12:53you won't see anyone outside of that, which certainly prevents spam, but at the cost of discoverability.
- 13:00We'll come back to how to design a more effective anti
- 13:04impersonation system later.
- 13:06Principle of relativity
- 13:08in cyberspace.
- 13:10The current web and our societies in general
- 13:14are built on an outdated paradigm that relies on and requires consensus about
- 13:19what is true,
- 13:21false,
- 13:22good, or bad.
- 13:24Unspeakable
- 13:25atrocities have been committed to achieve and maintain such consensus.
- 13:30However,
- 13:31we must move beyond this legacy of the past if we are to usher in a new era of freedom. The notion of a commonly shared belief or social truth is as dead in this new millennium
- 13:43as God was in the 19th century.
- 13:47The principle of relativity in cyberspace
- 13:50states, you cannot and should not dictate what is true or false,
- 13:55what is good content or bad content or spam
- 13:59for another person.
- 14:01If you do,
- 14:02you will simply be replacing the beast with a new set of beliefs.
- 14:07It's true that there are many different clients to choose from, but there were many different web browsers back in the day and that didn't end well.
- 14:16The reality is
- 14:18that a global agreement on things has seldom been necessary. In fact,
- 14:23relative canonicality
- 14:24can be achieved without the need for a central authority.
- 14:28Wikipedia, a Nostril alternative to Wikipedia, as you have probably guessed,
- 14:33fully embraces this idea of relative canonicality.
- 14:37Each topic contains multiple entries.
- 14:40There is no single source of truth that aims or claims to be neutral. Of course, this doesn't mean that all entities are equally good or useful. On the contrary,
- 14:51there is a very opinionated
- 14:53and biased way of sorting them, which depends on the observer.
- 14:57Regardless of the specific implementation,
- 15:00again, number on the forehead,
- 15:02this is the way to go.
- 15:04This is relativity
- 15:06in cyberspace.
- 15:08There is no global.
- 15:11The point is arguably a corollary
- 15:13of the previous principle.
- 15:15But it's so important that it needs to be fully explained.
- 15:19The notion of global is also dead.
- 15:21Despite being the largest index in the world, Google only indexes
- 15:26a very small portion of the web.
- 15:28The engineering challenges behind
- 15:31indexing absolutely every person and
- 15:34every piece of content in an ever growing open and hostile decentralized network
- 15:39will likely exacerbate this outcome.
- 15:41Nostr will outgrow
- 15:43every index.
- 15:45As spammers become more sophisticated,
- 15:48crawlers will have to become more selective,
- 15:51limiting the percentage of the network that is indexed.
- 15:54Therefore,
- 15:56a client designer or developer,
- 15:58you really only have a limited view of what's happening.
- 16:03This reinforces the statement that you cannot decide for someone else what is good or bad. True or false?
- 16:10You can't possibly know because
- 16:13what you have is a local limited view, just like everyone else's.
- 16:19Let's take Nosterband for example.
- 16:21As much as I appreciate their cool statistics, it's important to know their limitations.
- 16:27The version of trust rank that Nosterband
- 16:29implements,
- 16:30one,
- 16:32runs on the subset of the network that they are aware of.
- 16:362, uses an arbitrarily chosen set of NIP 0 5 providers
- 16:41for determining the so called good seeds.
- 16:45This approach is fine if it's used to defend the service provider against spammers. See DOS prevention section for improvements.
- 16:54However,
- 16:55such an arbitrary view,
- 16:57all are, but some less than others,
- 17:00shouldn't be imposed on the end user, but only used as a way for the service provider to select the customers it wants to serve.
- 17:10Principle of natural patterns.
- 17:13I believe that technology is made for people, so it should serve us
- 17:18and improve our quality of life. We shouldn't adapt to technology.
- 17:22Technology should adapt to us.
- 17:25This context is no exception to the rule.
- 17:29Humans are social creatures
- 17:31and our social dynamics and behaviors have evolved through our long history.
- 17:36Therefore, before introducing new and artificial social dynamics,
- 17:40it's wiser to mimic the natural dynamics that have evolved with us for millennia.
- 17:47The principle of natural patterns states,
- 17:50when in doubt,
- 17:52try to mimic the social dynamics of meat space.
- 17:56For example, in meat space, a friend of a friend is not a stranger,
- 18:01but it's not a friend either.
- 18:03It's something in between
- 18:05and a good candidate
- 18:07for a potential
- 18:08future friend.
- 18:10Another example is something that will hit home if you've ever been to a Bitcoin conference.
- 18:15When you're among people with the same interest,
- 18:18and I would argue mostly overlapping values,
- 18:21even if you've never met them,
- 18:24you can't say they're strangers.
- 18:26Examples abound,
- 18:28so there's really no reason to try to impose any artificial dynamics before exhausting what nature
- 18:34has given us.
- 18:36To summarize,
- 18:38we take from this section
- 18:403 principles
- 18:41that will inform
- 18:42our analysis and design decisions moving forward.
- 18:46The principle
- 18:47of least interference,
- 18:50the principle
- 18:51of relativity in cyberspace,
- 18:54The Principle of Natural Patterns
- 18:58The Problems
- 18:59Finally,
- 19:00we can start discussing some concrete problems and ways to solve them. There are several well defined problems that social graph analysis can help with. Again,
- 19:10it's important to define them well if we want to follow the principle of least interference.
- 19:17The first problem we will analyze is protection against impersonation.
- 19:22One of the most important pieces to get right.
- 19:27Anti impersonation,
- 19:29social endorsement.
- 19:31Who is the real Giacomo in this case? It's pretty simple.
- 19:36Pausing to describe a picture.
- 19:392 different Giacomo
- 19:41Zuco Twitter account
- 19:43banners.
- 19:45One states,
- 19:47has 68,000
- 19:49followers, the other says 44.
- 19:53First of all,
- 19:55let's talk about what won't help us decide.
- 19:59The name,
- 20:00the profile picture,
- 20:02the bio,
- 20:04the date of creation.
- 20:05And no,
- 20:07not even the number of followers as this is a completely irrelevant metric in an open and hostile
- 20:14network.
- 20:15What helps us is the line below.
- 20:18Quote,
- 20:19followed by Bitcoin Africa Story,
- 20:22Nostril Wallet Connect,
- 20:24and 666
- 20:26others
- 20:27you follow.
- 20:29This is the signal.
- 20:32Follows can be faked, but follows from people you trust cannot be faked
- 20:37or are more expensive to get.
- 20:40This is a particularly
- 20:42good design because it adheres to the principle of relativity in cyberspace.
- 20:47It doesn't prescribe
- 20:48who the real Giacomo is
- 20:51like the aforementioned
- 20:53WOT score where every follower has the same weight
- 20:57but it shows you clues that you can decide to interpret in the way that you want
- 21:03the same idea can be applied to mutes which give you a few more clues.
- 21:08Obviously, such a system can be extended not only
- 21:12to the people you follow,
- 21:13but also to the people who are 2 hops away from you or
- 21:18any number of hops away from you.
- 21:21However,
- 21:22the signal becomes increasingly diluted and unreliable the farther out you go.
- 21:28This approach is a good start but it can't really solve the impersonation problem on its own because
- 21:34it's limited in reach
- 21:36and suffers from a bootstrapping problem for new users which is really the same problem if you think about it
- 21:44There is also a subtle issue worth mentioning.
- 21:48What if a reputable person
- 21:50tries to impersonate someone else?
- 21:52This attack is possible
- 21:54because it won't be detected immediately.
- 21:57But the more prominent the person,
- 21:59the faster this attack will be detected
- 22:02causing the attacker
- 22:04to lose their hard earned reputation.
- 22:06Furthermore,
- 22:07for truly prominent figures such as heads of state,
- 22:11this problem can be mitigated
- 22:13but not completely solved by third party credentials
- 22:17that can be displayed by clients.
- 22:20NIP 0 5 Providers.
- 22:22Another approach that has been tried is to rely on NIP 0 5 Providers.
- 22:28If I trust a NIP 0 5 provider to verify the people they serve
- 22:33then I can assign a non zero level of trust to their
- 22:38However, this approach obviously requires trust which isn't ideal
- 22:43since there are trustless approaches that provide
- 22:46better
- 22:48guarantees.
- 22:49Proof of work keys.
- 22:52When we see a stranger enter the pub where we are, or when we are crowded on a subway, we find ourselves potentially
- 23:00vulnerable to strangers.
- 23:02In most civilized places, however, we don't live in a constant state of fear.
- 23:08And that comes from the fact that others are vulnerable as well.
- 23:12Everyone has skin in the game, even the aggressors,
- 23:15which means that the best way to play this game is not to mess with other people.
- 23:20On Noster however, the only consequence you can feel is the loss of your reputation,
- 23:26which by definition is worthless when the in pub is brand new
- 23:31however
- 23:32there is a way to add skin in the game for new keys
- 23:37For example,
- 23:38proof of work keys.
- 23:41The process works like this.
- 23:43While creating your identity,
- 23:45your machine can generate a new private key
- 23:48until the corresponding
- 23:49public key,
- 23:51as a binary number, satisfies the following inequality
- 23:56the public key is less than or equal to some threshold, T,
- 24:02of your choice.
- 24:04Since there is no efficient way to do this, allegedly,
- 24:07this process requires the expenditure
- 24:10of computing power and energy.
- 24:12This adds a quantifiable
- 24:15and verifiable
- 24:16cost to any identity
- 24:18making credible impersonation
- 24:20expensive.
- 24:22However
- 24:23proof of work keys can't work at scale because 1,
- 24:27they require commitment at key creation
- 24:30which is a bad user experience because it reverses the
- 24:35try before you buy rule
- 24:382,
- 24:39the proof of work can't be updated over time.
- 24:43And 3,
- 24:44it is not possible to delegate the work without delegating
- 24:48the key, supposedly.
- 24:50The last point implies that the economy of scale favors the attacker
- 24:55because they have multiple targets whereas the defender
- 24:58can only defend
- 25:00themself
- 25:01however there is a similar approach that offers what I believe are more appropriate trade offs
- 25:08proof of work
- 25:09endorsement
- 25:11Proof of work endorsement.
- 25:12A proof of work endorsement is a simple unsigned
- 25:16user metadata event containing a target in pub,
- 25:20a difficulty commitment threshold, t,
- 25:23and a nonce.
- 25:25The nonce is randomly generated
- 25:28until the following condition is satisfied.
- 25:31Some event
- 25:32is less than or equal to 2 raised to the t power
- 25:38where the event is a cryptographic
- 25:40hash function such as sha256.
- 25:44If the condition is met,
- 25:46the proof of work endorsement is valid and the target npubs
- 25:51powwait
- 25:53becomes
- 25:54256
- 25:55minus t.
- 25:57What I am describing is basically NIP13,
- 26:00except that it should use kind 0
- 26:03and it doesn't need a signature.
- 26:06What makes proof of work endorsements different from proof of work keys is that one, it doesn't require
- 26:12any commitment at key creation
- 26:15The proof of work can be updated over time.
- 26:18It is possible to delegate the work without delegating the key.
- 26:22And 4, requires data availability
- 26:26because the work is separate from the key
- 26:29point 3 is what makes this approach feasible and scalable
- 26:34especially for mobile devices constrained by CPU
- 26:37and battery limitations.
- 26:40Point 4 isn't particularly bad because the events are small,
- 26:43public and the user has an incentive to store them.
- 26:48Example of how it works.
- 26:51Alice
- 26:52gives a n pub to the miner.
- 26:56Alice pays the miner $10
- 26:58The miner then mines the proof of work endorsement
- 27:02with the target n pub.
- 27:04The target n pub now has a non zero proof of work wait.
- 27:10How does it help?
- 27:11Well, if the target impub has a non zero proof of work weight,
- 27:16credible impersonation attempts require a cost.
- 27:20The attacker's
- 27:21social capital or proof of work weight becomes less and less valuable and is eventually lost as more people discover the attack and mute the attackers in
- 27:31pub. Furthermore,
- 27:33assuming that most of the proof of work weight is allocated
- 27:36to real people,
- 27:38it's possible to use its distribution to fix the shortcomings
- 27:42of social endorsement, for example.
- 27:45A new user
- 27:47can be presented with the choice of following the inpubs with the highest
- 27:51trust rank
- 27:52where the good seeds are the inpubs with non zero proof of work wait.
- 27:58The same principle can be applied to an existing user trying to get information
- 28:02about the inpub that is many hops away from him or her.
- 28:06If you want to summarize this idea to the maximum,
- 28:10if the number of followers is meaningless
- 28:13that's not true for the number of followers
- 28:15with non zero proof of work weight
- 28:20I have started working on this idea and preliminary results are promising, but more thought is needed
- 28:27Colored
- 28:28Halo.
- 28:30Another anti impersonation mechanism
- 28:33tried by Nostrudel
- 28:35is the use of a few characters of the public key in hex format
- 28:40to identify
- 28:41a unique color that is displayed around the user's profile picture.
- 28:46The idea is
- 28:47that if I am already following the real Lynn Alden,
- 28:51a credible impersonation attempt would require generating
- 28:55many times the key pair
- 28:57to get the same or very similar color.
- 29:01That requires energy which means
- 29:03cost.
- 29:04However,
- 29:05this information is only useful if I already follow the real Lynn and remember her color
- 29:11asking my trusted network for the color of a particular person is no different task than asking her in pub directly
- 29:19This means that this approach, while interesting,
- 29:23is not very useful.
- 29:25Denial of Service Attack Prevention
- 29:29When analyzing denial of service attacks,
- 29:32we should distinguish between defense mechanisms that work at the network level,
- 29:37at the service provider level,
- 29:39and at the user level.
- 29:42Talking about the first case may seem to contradict the principle of relativity in cyberspace
- 29:47but here by network we mean any set of network participants
- 29:53who are willing to cooperate but not necessarily
- 29:57trust each other
- 30:00Network Level
- 30:02proof of work events basically apply the concept
- 30:05of HashCash
- 30:06not to email messages but to notes.
- 30:09There is not much to say here except to note that the proof of work can be outsourced
- 30:14to service providers that can most likely out compete the attackers inefficiency.
- 30:20Service Provider Level.
- 30:22The best tool the service provider can use to prevent DOS' money.
- 30:28A paid service, either pay per use or subscription base,
- 30:32can work both for monetization
- 30:35and as a defense mechanism. However,
- 30:38not every service or sub service is directly monetizable,
- 30:42which is where social graph analysis can help.
- 30:45The idea is to selectively choose which end pubs are considered potential good customers and filter out everyone else.
- 30:54If this needs to be said, I'll say it. This is not censorship.
- 30:58This is simply a business deciding whom to serve.
- 31:02Many different criteria can be used to select what is a good customer, especially since this definition depends
- 31:09on the service being provided.
- 31:11A general approach is to use trust rank or spam mass estimates
- 31:17with a suitable measure for selecting the good seeds such as proof of work weight
- 31:21as the criteria for deciding the
- 31:24good customers.
- 31:26Those who don't qualify will simply not be served for free.
- 31:31User Level.
- 31:32Here we consider a simple defense against DoS attacks on user DMs.
- 31:39The user has a whitelist of end pubs that can directly DM him or her for free.
- 31:45If the sender is not on the white list, the DM
- 31:49will only be displayed by the client if it is preceded by or comes with a payment above
- 31:56some certain threshold.
- 31:58The white list can be edited manually, but it's automatically
- 32:01populated
- 32:02based on some social graph criteria that the user has approved
- 32:07for example,
- 32:08all inpubs in some set can be automatically added to the white list of user v.
- 32:16Personalized ratings.
- 32:18The idea that every opinion or vote is equally important to everyone is a byproduct of political democracy
- 32:24and its sanctification.
- 32:26This is never the case in any social context.
- 32:30Imagine you want to buy a product that has thousands of good reviews
- 32:35but your friend
- 32:36recently bought it and tells you that it's not good.
- 32:39It's natural for you to give more weight to your friend's opinion contrary to the belief that each opinion has equal weight.
- 32:47When computing an average
- 32:49there is always an assumption about the weight distribution
- 32:53for example, how important
- 32:55each vote is
- 32:57Social graph analysis can be useful to design more accurate and even personalized ratings. For example,
- 33:02ratings that reflect who the viewer is
- 33:06and what he or she is looking for at any given time. For example, you might want to give more weight to people who have similar interest to you or more weight to your friends and friends of friends
- 33:18many different criteria for calculating weight distributions are possible
- 33:22making this an exciting area for future research
- 33:26it's important to note
- 33:27that the most trustworthy ratings will be those that are independently
- 33:31verifiable
- 33:32both in the ratings
- 33:34and in weight distributions.
- 33:36Practically speaking,
- 33:38there are 2 steps involved in computing a weight distribution
- 33:42select the support
- 33:44for example, the set of end pubs that have a non zero weight
- 33:48And 2, assigning a specific weight to each end pub that is part of the support.
- 33:55These two steps can be done simultaneously
- 33:58or one after the other. In the next section, we'll look at support selection.
- 34:03A problem that arises in another seemingly unrelated context,
- 34:07social discovery.
- 34:09Social discovery.
- 34:10How many times have you discovered something thanks to a friend's recommendation?
- 34:15This phenomenon can actually be productized
- 34:18to streamline the discovery of content, places, and people that are relevant
- 34:23to the end user. The goal is to construct another graph
- 34:27consisting of the recommendations for each of the users we want to serve.
- 34:32This is a simplified scenario
- 34:34where what is being recommended
- 34:36are other nodes
- 34:38in the graph.
- 34:40Now, I need to pause here to tell you that this next section
- 34:45is essentially
- 34:46nothing but
- 34:47matrix
- 34:48math.
- 34:50Lots of equations.
- 34:52It's almost impossible to read
- 34:55in the format that I'm reading it to you now. So I'm going to essentially
- 34:59describe
- 35:00what it is that we're looking for here. I'm going to go down here to this one picture
- 35:06that has 3 different kinds
- 35:09of
- 35:11results of matrices.
- 35:13Right? And what what he's wanting to do
- 35:16is, under certain circumstances,
- 35:19who would you most likely follow or who is a good candidate to follow or who are you connected to that you might not know you're connected to? Like, for instance,
- 35:29there's a notion that he gives of direct propagation.
- 35:34Like and let let's call the the vertices
- 35:38users.
- 35:39Right? So there's user 1,
- 35:41user 2,
- 35:42user 3, and they're connected in some way and the users that are the nodes,
- 35:48each one of us is a node.
- 35:50So if I'm a user and I'm connected to another user, say user
- 35:552, and that
- 35:57person is connected to user 3, but I'm not directly connected to user 3,
- 36:02I have a direct
- 36:05propagation
- 36:06over to user 3
- 36:08through my relationship
- 36:09with user 2.
- 36:11There's another type called
- 36:13transposition
- 36:14and this one I really like.
- 36:16It's the notion that if there's 3 users
- 36:20and I'm connected to user 2
- 36:22and a third user is also connected to user 2
- 36:27and we're both following the same person. In the first example, I was following user 2 and user 2 was following user 3.
- 36:34In this
- 36:35case both I, user 1, and user 3 are following
- 36:40user 2
- 36:43So user 1, me, and user
- 36:463, who I have no direct connection with,
- 36:49are finding something that is of value in user 2. So therefore,
- 36:54we have
- 36:55kind of a connection
- 36:57and we can exploit not exploit that. We can
- 37:00use that,
- 37:02and it's called
- 37:03transposition.
- 37:04The first
- 37:05one is called direct propagation.
- 37:08The third one is very hard to describe. It's called
- 37:11co citation.
- 37:13So
- 37:14me, user 1,
- 37:16is directly following user 2
- 37:19and user 3.
- 37:22User 4
- 37:23is directly
- 37:24following
- 37:25only
- 37:26user 3
- 37:28but
- 37:29because I
- 37:30am following
- 37:31user
- 37:333
- 37:34and user 2
- 37:36and user 4 is also following like I am user 3,
- 37:41then user 4
- 37:43has some
- 37:44not ephemeral, but like let's call it an ethereal connection
- 37:49to user 2.
- 37:50The assumption is because I'm following user 3 and I'm following user 2
- 37:56and because user 4 is following user 3, then user 4 might also be interested
- 38:01in what user 2
- 38:03has to say
- 38:05because I'm following
- 38:07the same person that user 4 is following. So even though user 2 isn't directly connected
- 38:12in that relationship,
- 38:15There is an ethereal connection
- 38:18simply because both me and user 4 are connected to user 3. So therefore, because I'm connected to user 2 and user 4 isn't
- 38:26user 4 may very well be interested
- 38:30in what user 2 is trying to say.
- 38:33By combining
- 38:34and experimenting with propagation rules and weighting criteria,
- 38:38it is possible to efficiently compute personalized,
- 38:42verifiable, and dynamic ratings for each user,
- 38:45making this an exciting area for future development.
- 38:49Conclusion.
- 38:50This is a comprehensive
- 38:52yet superficial paper as it touches on many different areas without exhausting any.
- 38:58This was done intentionally
- 38:59to stimulate interest in the field of social graph analysis.
- 39:04More importantly,
- 39:05it aims to help the reader understand the significant impact this exciting new field can have on the web, although not without
- 39:13its risks.
- 39:26The Circle P is open for business and today's vendor is the Leathermint
- 39:30dotcom. That is the leathermint.com
- 39:33if you need a belt, if you need a
- 39:37passport holder, if you need a a a specifically a wallet, and some of the coolest wallets that there are, you will go to the leathermint.com,
- 39:45or you can find them on Noster at leathermint
- 39:48or on dead bird site formerly known as Twitter at
- 39:53the leathermint.
- 39:54And if you use their coupon code and type in
- 39:57bitcoin and
- 39:59all one word, you'll get 10% off of your purchase, and they will return to me
- 40:04some satoshis for putting them on the air. Honestly,
- 40:08some of the best leather I've ever seen, some of the best craftsmanship I've ever seen, and as I always continuously harp, you cannot
- 40:17not not
- 40:18look at the stitching.
- 40:19If your stitching fails, the entire wallet fails.
- 40:23So if you want some quality, if you want sovereignty in your style,
- 40:28go to the leathermint.com.
- 40:41Okay. So I don't want to belabor this too much longer, but there's one particular thing
- 40:47in this piece that I I'm just kinda fascinated with. Now
- 40:52he kinda leans a little too hard on web of trust, the WOT scores
- 40:57as given in Coracle.
- 40:59I think that those are fine. I I really do. I understand
- 41:03his arguments that suggest that it's not optimal.
- 41:07But just because something isn't optimal,
- 41:09doesn't necessarily mean that we need to throw the baby out with the bathwater.
- 41:14I like the web of trust idea. I think that we haven't seen it blossom yet.
- 41:19However,
- 41:21this particular gentleman's idea of proof of work keys,
- 41:25I cannot get out of my head.
- 41:28This makes sense to me. And I also think,
- 41:31honestly, in a very real way,
- 41:33I think that there's no reason that this couldn't work very well
- 41:37alongside
- 41:38web of trust or the w o t score
- 41:41in order for us to really filter out
- 41:44the garbage.
- 41:46We need something to take out the trash.
- 41:48It doesn't matter what you're doing. If you're just if you're cooking dinner, you're going to generate trash.
- 41:53If you have old clothes at one point or another
- 41:56you're going to call those and you're going to have to throw them in the trash because you can't even donate them. They got like some of my t shirts have huge holes in them. I'm not going to donate that. That's ridiculous.
- 42:06You know, I use them outside
- 42:08mowing the lawn.
- 42:09But no matter what you do, you're gonna generate trash.
- 42:13You're going
- 42:15talk to a programmer
- 42:17and ask them about how they handle their trash
- 42:19in in coding, and they will have a solution that they use to handle garbage,
- 42:26whether it's what they do to keep their code clean or what the compiler does. It does
- 42:31something is going to do trash handling.
- 42:34Right? And we have it in social media as well. And not just social media because Nostra's much more than that, as I've said before. But
- 42:43no matter what you do, there's garbage.
- 42:46And this this issue of of spinning up this
- 42:51public private key pair
- 42:54but it has to be mined
- 42:56I like this idea because this would stop
- 43:00the issue that I was telling you about before we began reading the piece.
- 43:05I'm getting bombarded.
- 43:06I could go to primal right now. I was there about an hour ago, and there's probably 27 new followers
- 43:13and only 3 of them are somebody that I might actually care about.
- 43:183 of them that I might consider
- 43:20human because they have, you know, a a different avatar
- 43:24than the avatar
- 43:26style that's coming down with these other 20, what, 24
- 43:31followers. They all have these orange concentric rings that are like a clock, and it's they're all in a different confirmation, and it's really weird. And
- 43:39it's clearly not a human. I mean, so right there, I'm I'm I've got an algorithm in my head saying I don't want any of these guys. I want the 3 guys that I'm I'm pretty sure are human. Doesn't mean that they are. It just means that there's more of a chance that they're human.
- 43:55But what would even be more of a chance than that
- 43:59is if they had real money to put on a line
- 44:03to build that private key.
- 44:05And say look I want a nostril key okay you get your first one for free
- 44:09and there's there's no way that you can really
- 44:12do what I'm suggesting here, but just bear with me.
- 44:16You get your first one for free. Everything else you gotta buy.
- 44:20You know, you you gotta you or or you could spin them up, but I'm I I've got a setting in my client that says unless you paid
- 44:29at least a dollar
- 44:31at least 2,000 satoshis. Unless you pay 2,000 satoshis
- 44:35for your proof of work key, which I would be able to tell by how much proof of work is
- 44:41tied and forever tied to your key,
- 44:44I I you you get filtered out of my feed I don't even see you you can follow me I'll never know it it doesn't matter and also, by the way, I might actually have a setting that I can automatically
- 44:55mute you. And that way, the web of trust could pick up, oh, you've muted this person. Doesn't count to your web of trust score. Oh, and by the way,
- 45:04huddlebond, if,
- 45:05I need to have a discussion with you a lot more a lot longer about web of trust and how that calculation works.
- 45:12And what happens? Here's my question to hodelbod.
- 45:15What happens
- 45:16if I have
- 45:183,000 followers out of 12,000
- 45:20followers that are obvious fakes and nobody else gives a shit about? Nobody else can even see them because they're only following these 17 people and they're not engaging with the community at all. Does that affect my web of trust? So that's my question directly to Huddlebot.
- 45:37I think that it kinda could, but I'm not sure, so I'll wait for the answer.
- 45:42But I don't believe for an instant
- 45:45that there's no way that Web of Trust
- 45:47doesn't have value. It does.
- 45:49But I also believe that in conjunction with
- 45:53having to spend some satoshis,
- 45:561,000,
- 45:572,000,
- 45:5910,000,
- 45:59how much should it cost? I don't know. I
- 46:02no idea whatsoever.
- 46:04That's not the point.
- 46:06The point is that
- 46:08for me to be able to spin up something on like a new account for something,
- 46:13I would have to actually
- 46:15do things
- 46:16and reach for stuff and and solve particular problems
- 46:20that we would like to believe that only a human could solve, to jump through
- 46:25at least a couple of hoops.
- 46:27I love the fact that it's easy to spin up a key on Noster, but I don't like the fact that, you know, a full one fifth to one quarter of my followers are clearly fake.
- 46:37They just are. I don't I'm not gonna like, at this point, I don't even wanna say, oh, I got 12,000 followers, so they're you know, I'm I'm badass. No.
- 46:46Not when a quarter of them are obvious fakes.
- 46:49So
- 46:50we have issues.
- 46:51We're always gonna have these issues. Nothing is ever going to be solved because
- 46:56if everything was solvable, everything would already be solved.
- 46:59No new problems would crop up and we could just all go to our graves happy as a clam because there's no reason to actually live anymore. You see what I'm what I'm getting at?
- 47:10But these problems to me are fascinating
- 47:12and I hope you enjoyed this piece.
- 47:15Please go follow this Papelia guy.
- 47:18I'm hoping,
- 47:20that he has some more stuff about this coming on in the future. He has a very interesting way of publishing.
- 47:27It has something to do with,
- 47:29my favorite note taking app, Obsidian. I'm not exactly sure how he's doing it. Papalia,
- 47:34if you're listening,
- 47:35please answer my nostril note asking you the question exactly how you're leveraging Obsidian
- 47:41to make these notes or these, publications because it's honestly,
- 47:45I'm fascinated.
- 47:47Can't get my head to not stop thinking about it, so
- 47:51get in touch with me. Tell me how you did it, and I will see you on the other side.
- 47:55This has been Bitcoin and and I'm your host, David Bennett. I hope you enjoyed today's episode and hope to see you again real soon.
- 48:04Have a great day.