debrief33

n at
0:05the ascribe offices for about half the
0:09week met did an intensive kind of a
0:14presentation and then discussion with
0:17Trent Makana he and his team and then
0:23did a deep dive with their architects to
0:25see you know where were their possible
0:28points of collaboration so quite quite
0:31excited about that I think it was you
0:33know there was a lot of interest and
0:36there's a lot of it's a good timing from
0:40the point of view that they're they're
0:42kind of doing an architectural review
0:44planning for their next phase of the
0:48next iteration so it was very very
0:51fruitful in that way let me I think
0:54there might be some practical ways that
0:58we can help each other out I'd love to
1:00see the Archon node with a sort of the
1:052.0 version of the big chain DB
1:08underneath as a data source so that
1:11could be quite interesting also I met
1:15with Clement
1:17FBA who is at the focal point of a very
1:23interesting group in Paris
1:26that brings together startups and and
1:30research groups and and sort of in the
1:34incumbent industry to to do a practical
1:40collaboration and explore new ideas he
1:44just had one of his projects was a
1:48blockchain energy project called dayz
1:53that just came back from doing some
1:57field work to try to explore the
2:00feasibility of doing doing blockchain
2:06energy projects in the developing world
2:13so I'm quite interested in that I'm very
2:15interested in making sure that our chain
2:19is vigorously pursuing all avenues with
2:23respect to blockchain and decentralized
2:27energy solutions so that was very
2:30exciting
2:30there are three points where we are
2:33looking to collaborate and one of the
2:36most important is that he needs
2:38artifacts from arch tang to begin to
2:46socialize the ideas of our chain which
2:50is why it's it's such great news that
2:52Kant wanted me to pass this along so we
2:55now have the full ERC 20 token con
3:01contract compiling in rolling and
3:08executing on top of the rosette p.m. so
3:12this is a major major milestone and
3:14that's been checked in
3:16in addition looks like Kent and navneet
3:22and Glynn did some work and got a
3:25rolling Oryx site up and running so it's
3:30not quite a rebel it's still very
3:31nascent work but this is exciting
3:33because we're at a point now where we
3:36have a little web-based experience where
3:39developers can come and type in row line
3:43code and have it evaluate and see what
3:45the results are so that's that that is
3:47ongoing work it's still very rough
3:50around the edges but I'm excited to see
3:52it come into being so this is this is
3:54really great additionally we've I've
4:01been pursuing with higher effects the
4:05idea of sort of taking on board the
4:11development of development tools and and
4:14developed an IDE and relations with with
4:17the incumbent IDs and they've sort of
4:20been some positive response around that
4:22so so more to come on that
4:25as the ideas develop but my aim is
4:30however how however and whoever takes it
4:34on my aim is that as soon as mercury is
4:38out the door right on the heels of
4:39mercury we have a bunch of developer
4:41tools for for rolling and for our chain
4:45in general so this is this is part of
4:48the plan and we're very very excited to
4:51get this underway as soon as possible
4:54also since I've been here in Berlin I
4:57met with with Tim and we went over a few
5:01questions about the the parser and I Tim
5:06also joined the conversation with
5:08Clement and I thought that was really
5:11important because we want to make sure
5:13that you know everybody you know has a
5:16chance to everybody understands kind of
5:19the bigger picture the the larger goal
5:21our goals and and you know kind of knows
5:27how their their pieces fit into the fit
5:30into the efforts that we're doing so it
5:32was really good to meet with Tim it's
5:35Tim on the call yes I just send you an
5:46email regarding further pasta questions
5:49and besides it was nice meeting you was
5:53a good talk oh yeah it was also
6:01interesting I think Tim had a really
6:04nice attitude
6:05it couldn't quite make be NFC work for
6:08some of the rosette parser stuff so he
6:12went down the path using antler and was
6:14able to get get something going and I
6:17think that's that's kind of the way we
6:18we need to be not necessarily wedded to
6:21any one platform as long as we get the
6:24stuff done so I was very excited to see
6:26that and then finally I started to
6:33sketch out a framework or
6:39testing caspere which is a bit one of my
6:42goals to make sure that we have you know
6:45something that's you know a simple test
6:49and yet compelling and worthwhile
6:52building a simple kind of test remember
6:55that begins not only test just the code
6:59but also tests some of the implications
7:03and I will I'll go through those ideas a
7:08little bit later but I wanted to pass
7:12the baton over to Edie and and see if he
7:16has any updates on the holding side and
7:19and just in the updates in general hey
7:22hi everyone
7:25yeah we've we've had a number of
7:28investor conversations that are
7:30interesting you know one one guy met
7:34with this morning here at the navneet
7:37and I met at the Redmond office he he
7:40literally heard about us at an airport
7:42someone was talking to him about
7:43blockchain in our chain and he lives in
7:46the Seattle area and was so excited and
7:48wanted to meet us he heard about in
7:51particular this the safe and so he came
7:53in and just across some conversation so
7:55like now our train is an airport really
8:02really interesting exactly last Thursday
8:13Lisa and I met with a sort of a private
8:19investment / Business Network group here
8:23in the Seattle area as well I was
8:27invited by this group because they
8:29wanted to understand about blockchain
8:31and you know this was a diverse group
8:35they were folks ranging from an
8:39executive at a local bank to insurance
8:43brokers and there was an M&A guy there
8:46was an attorney who's working on a token
8:49sale for a Seattle area
8:52business and so it was is quite an
8:55interesting conversation just walking
8:58through all the misperceptions that
9:00people have about bought chains and so
9:02that was a very great education session
9:05and then after that met with this M&A
9:07guy for an hour talking about our chain
9:10and it was great getting his initial
9:12impressions and and so forth to to in
9:17particular that we can talk more about
9:20it some some later time was he had
9:22concerns about the supply of of staking
9:25tokens and the governance around that
9:28supply and use of the stake token so
9:30Gregg and I talked yesterday morning
9:32some about that how we can make that
9:36more clear on the co-op side and also
9:39make some decisions board level
9:41decisions as opposed to individual
9:43decisions and that will help give
9:45investors confidence such as this guy
9:48who's considering an investment in the
9:51holding company so anyway that was an
9:53awesome experience and conversation as
9:58as Greg mentioned last week the coop is
10:02preparing for a private sale of rocks
10:04and we've retained two advisors that are
10:08associated with david otto and i've met
10:12with them and are continuing to work
10:14through sort of materials that are that
10:17are needed to better understand the
10:20co-operative and the opportunity in
10:22front of us all
10:23Greg's planning on writing a white paper
10:25to help that effort these these advisors
10:29and and us they will write a business
10:34summary that will include a financial
10:37financial forecast especially from the
10:40expense side and then also I'm going to
10:43take a another pass at the architecture
10:45document just because I haven't looked
10:47at it since like January and and so
10:52understanding and so forth has changed
10:55we'll have a working session later this
10:58week on this business summary as well
11:02also I've been talking to
11:06and with Nathan about our marketing and
11:09brand image is Nathan on okay so he's
11:15not he's not on our intention is to
11:19write a branding document about you know
11:23what and this will help also with the
11:26investor pitches you know what are the
11:30the primary you know ethos and messages
11:34that we want to change as our chain from
11:36a our sorry uh convey as as our Jane
11:39cooperative and what kinds of feelings
11:43do we want to stir up in our messaging
11:45on marketing and then that will lead to
11:47a discussion around the logo for the Rev
11:50token and maybe the rock token if we
11:54choose to have a separate logo for that
11:57I'm expecting Nathan will drive a lot of
12:00this discussion and OB with with a lot
12:03of input from the community I was hoping
12:05you'd be here to to discuss that but
12:07first I'd like him to do to have this
12:10conversation around branding so we
12:13started some discussion so to I think
12:16that's all I have
12:19sweet no thank thank you very much that
12:23was really good
12:25also I Jim White's cover had some
12:30interesting comments regarding that some
12:32of the stuff that I discussed and I
12:35wanted to make a couple of clarifying
12:37points so recently and I want to keep
12:40harping on this what I have said is that
12:43that if the the Kaspar algorithm works
12:47as we expect it to and I have every
12:50confidence that it does then one of the
12:53one of the things that we'll be able to
12:56do is to replicate arbitrary rewrite
12:58systems you know whether you rewrite
13:01system is JavaScript or or what-have-you
13:03and what this does is commoditized a
13:06platform that doesn't mean that it makes
13:08the platform not matter the platform
13:12still matters but what it will do is to
13:15greatly highlight
13:19Adam berate and make clear that the
13:24differences in the computational models
13:27so in particular if you have us you know
13:30like you you should be able to replicate
13:32regular expressions but our regular
13:34expression is going to be you know the
13:37kind of computational model that fits
13:39this domain no they aren't and they
13:41clearly aren't and and anyone who who
13:43gives it you know two seconds where it's
13:45a thought will recognize regular
13:47expressions are not well suited to this
13:48domain so but but you can still use the
13:53the replication algorithm consensus
13:57algorithm to to replicate a regular
13:59expression execution across a bunch of
14:02decentralized nodes so you know and
14:05there might even be value in that but
14:07it's just it's not going to be it's not
14:11going to meet a lot of the the demands
14:13that we've been talking about or that
14:15have emerged as as people consider the
14:17space the likewise any sequential model
14:21by comparison to a model that is
14:25inherently concurrent right replicating
14:28those different kinds of computational
14:30models who will make the features of the
14:34computational models very very starkly
14:37clear and and that's what I was saying
14:40so the monetization of the platform
14:43means that a lot of what has gone into
14:47the capital requirements and operational
14:50requirements goes away which means the
14:54the the the MA the the the platform
14:57choices with respect to computational
15:00power computational fit for the domain
15:03all of those will come to the forefront
15:05and some of these other considerations
15:08will go to the back and that that's
15:12that's what I meant when I talked about
15:14the commoditization of the platform so
15:17so a lot of the value in the platform
15:20that sort of the heavy duty operational
15:23and capitalization costs will no longer
15:26be justified because you can press a
15:29button and get it done however
15:33differences in choices design choices
15:36around things like where they're using
15:37something that is fundamentally
15:39concurrent versus something that is
15:41fundamentally sequential those will come
15:43to the fore and that that's where that
15:47that's where we're we're you know
15:49healthy competition and cooperation will
15:54will you know really begin to ferret out
15:57that the best choice is the best design
15:59choices
16:00so I really appreciate Jim for sort of
16:03highlighting that conversation yeah
16:06taking you slightly different direction
16:08I guess I was in the context of using
16:16something like react the value keystore
16:24that's an excellent choice
16:26absolutely you can use any of a number
16:28of different key value stores react is a
16:31perfectly reasonable choice for a number
16:34of different perspectives you know I was
16:37thinking about like their search
16:41applications I was thinking that know
16:51that ro Lang program is sort of search
16:55application on top of react seem like
17:00it's it's possible
17:12we lost your Greg I'm still here can you
17:16guys hear me yeah we can now okay great
17:20so so you know I absolutely agree we
17:24should have lots of different
17:25realizations of our chain nodes on
17:29different kinds of stores and we'll sort
17:32of a post Mercury will lay out a roadmap
17:36for different kinds of storage and a
17:39plug-and-play architecture for different
17:41kinds of storage but but up to mercury
17:46we'll be focusing on one storage choice
17:50so unless there are any other major
17:55updates that we want to give I'd like to
17:58talk a little bit about this test
18:01framework and and also this whole idea
18:05of beginning to grow a test organization
18:09that that interfaces nicely with the
18:13community because we really really will
18:16rely heavily on the community for being
18:19able to you know to hold our feet to the
18:26fire and make sure that we have the
18:28right right quality bar so and and and
18:33that also and and I've organized the
18:35discussion here in a way that it also it
18:38sort of has this practical side which is
18:41you know a real test of consensus and
18:46then and then the other side is that it
18:49testing the ideas of consensus and what
18:52this means what the implications are you
18:55know what what does what does
18:56decentralization mean in terms of the
18:59existing infrastructure the existing
19:02compute and services infrastructure and
19:05I wouldn't lead I want to lead the idea
19:09lead off the discussion with the
19:12following you know you know Rick
19:15recounting or recognition of certain
19:17kinds of experience I think everybody is
19:20familiar with the experience of the
19:24conversation between the in
19:26dividual agent you know such as you know
19:28yourself or myself trying to talk to an
19:31organizational agents like a bank or a
19:35hospital or a government service and and
19:38we know what that experience can be like
19:40you know it's like you you you call into
19:43a line there's a phone tree that's the a
19:45phone menu that's about 20 levels deeper
19:47and at the end of that one is connected
19:51to another person and that person is not
19:54empowered to to take appropriate action
19:59nor are they nor are they actually
20:01informed of any of the information that
20:03was supplied in the the 20 level deep
20:07phone menu and so there's a
20:10recapitulation of the information and
20:12then there's a then there's the
20:13navigation of the the human
20:16organizational tree in order to get to
20:17someone who is actually empowered to to
20:20help with a with a question or make a
20:23decision or something like that so it's
20:25a very common experience and and the
20:29spies I'm going to present challenge
20:32that that's the way it has to be and and
20:36I'm going to say why that's important
20:38sort of the social context for why
20:42that's important as a part of this
20:44discussion so again I'm attempting to
20:47work on multiple levels all at once here
20:52so I'm screen sharing now and let me
20:57bring up the slide
21:01hang on a second so this is a
21:06aft and this draft will also a lot of
21:10the ideas and concepts of this draft
21:12will find their way into the white paper
21:14so I apologize that OmniGraffle has
21:20changed the way they do page layout and
21:22so that the presentation the slides the
21:25the the look and feel is not as pretty
21:27as it has been so kind of as I said
21:33before there there's several desiderata
21:35here so first we want to provide a kind
21:37of compelling test of the consensus
21:41algorithm and the second is to
21:43illustrate the nature of different kinds
21:45of agency and you know the interesting
21:49point is that different levels and
21:50different kinds of agency actually can
21:52can have as opposed to what we often
21:54experience in the world they can have
21:56fruitful and even satisfy in engagement
21:59and interaction so so the individual
22:03talking to the organization can be a
22:06completely different experience with a
22:09consensus algorithm and that's what
22:11that's one of the ways in which
22:13consensus has the potential to
22:16completely change a modern society so so
22:22I'm going to suggest that we build an
22:25organizational agency that can act as a
22:28single agent that this is in fact what
22:31consensus provides it makes it possible
22:34to have an organization so a bunch of
22:38clients acting as a single agent and in
22:43this case I'm gonna suggest that we make
22:45it since we're trying to build a test
22:49that we can write code to make this
22:52single agent one that plays a game
22:55gamification is a theme that runs
22:57throughout a lot of the work that I do
23:00large largely because we learn best when
23:03we play but so here what we're talking
23:07about as a community agents that act as
23:09though as a single agent playing as a
23:14player in a game and I'm gonna restrict
23:17attention just to two person games just
23:19for now
23:20like chess or go and the important point
23:24is that the organizational agency can be
23:27an aggregated AI or it can literally be
23:31a community of human agents who act as a
23:33single agent and that's sort of the the
23:36other the point that I was making it at
23:38the outset of this discussion so so and
23:42and this has implications for AI and you
23:45know what has been become you know a
23:48real there's there's a whole there's a
23:52lot of hoopla and hype around deep
23:54learning and I'm gonna make some
23:56proposals that that suggest that that we
24:00can do better than what is currently
24:02thought of as deep learning and that
24:04consensus plays an important part in
24:06that right and this is all within the
24:08context of just building a test to test
24:10Kasper so so the idea is that we want to
24:16we want to take say white player in a
24:18game of chess and it's made up of a
24:21community of client proposers and a
24:24network of validators and then the
24:27community is subdivided into little sub
24:29communities so here's here's a little
24:32picture and and the thought is that each
24:40sub community is proposing moves for
24:42only one piece right so you can imagine
24:46that you've got a bunch of clients that
24:48are moving one of the white pawns and
24:51another bunch of clients that are you
24:54know moving another of the white pawns
24:56and a bunch of clients that are only
24:58interested in moving the white rook
25:00right and so they're each proposing
25:02moves for that one piece and those
25:05proposals are headed in to Casper which
25:09is going to create a consensus around
25:12one move the D one move so again here's
25:17a picture that to reiterate this so
25:20you've got a bunch of clients and
25:21there's a single pawn and there are a
25:24bunch of possible moves for this pawn
25:27all right now one one would be just to
25:30advance forward one another would be to
25:32advance forward to
25:34and another might be to take a piece on
25:37the adjacent square and each of the
25:39clients is is potentially proposing a
25:42single transaction for the pawn for that
25:45pawn resource similarly for the night
25:48each of the client assisted proposing a
25:51single transaction which mutates the
25:53resource which is the knight right and
25:57now intriguingly if we restrict the
26:00block size to exactly one this
26:03guarantees that the validators will pick
26:05a single move from all the client
26:08requests and what that does is it allows
26:11the players to two alternate moves so
26:15white player can move and then black
26:17player can move and then white player
26:18can move and then black player can move
26:20and and what you're getting now is a
26:24dialogue between organizational agency
26:27and singular or individual agency under
26:29the assumption that black player is an
26:32individual agent and that assumption is
26:34actually worth looking into which and
26:40I'll get to that in a moment but but
26:42clearly if the consensus algorithm
26:45performs in such a way that black player
26:48if it's a human being doesn't get bored
26:50or timeout waiting for white player then
26:53you have an algorithm that's responsive
26:55it's responsive on the level that we
26:58kind of want to see consensus being
27:00responsive I actually want to see we
27:03want to see it massively more responsive
27:05but this is kind of a first cut at the
27:08level of responsiveness that we want to
27:10see and and more intriguingly
27:15if we stipulate that the black player is
27:18a single agency rather than an
27:20organizational agency then the black
27:24player may be interested in whether or
27:27not white player actually is an
27:31organizational agency or if it's a bunch
27:33of bots the controlled by one one entity
27:37right and so black player can actually
27:41demand a proof of consensus and and at
27:47least in
27:47a number of the the proposals that have
27:53come back and forth between GLAAD and I
27:55we have such a proof and the proof is is
27:58a trace or it's a set of sequences of
28:01justified messages that gives rise to
28:03the consensus decision so whenever you
28:05if any of you have watched the Casper
28:08hangouts when Vlad has run his his
28:11simulations for the for the the the
28:17ideal adversary the the message flows
28:21that reach the consensus decision that's
28:24exactly the proof of consensus and
28:28that's what white player can provide if
28:32it is an organizational player now it
28:36comes computationally expensive and
28:41economically expensive to fake this
28:43because every every validator will be
28:47staked and they would have to so you
28:50would have the the validators and there
28:53are stakes and all the clients and all
28:56of the transactions that were associated
28:58with the production of the consensus so
29:01that's that both the computational cost
29:03and the economic cost to fake that
29:07consensus so it becomes prohibitively
29:10expensive to fake proof of consensus so
29:13this is a fairly reliable way by which
29:18black player can ask that a white player
29:22is actually an organizational agent now
29:26it also has a separate role or a
29:29separate meaning which is that it stands
29:31as an explanation of the decision so the
29:35set of messages that go back and forth
29:38essentially provides a justification for
29:41the consensus which is a justification
29:44for the decision and therefore unlike so
29:48you can think about this so the
29:50consensus is very very much like a kind
29:52of the kind of flow you might see in a
29:54neural net except that the flow here
29:57corresponds to an explanation a
29:59justification so
30:01unlike a neural net if this were a bunch
30:05of eyes that were reaching consensus
30:08this AI that organizationally I can
30:11actually explain its move so that's
30:15quite that's quite intriguing when you
30:18compare it to deep learning and in fact
30:21the whole architecture as an
30:23architecture for AI is very similar to
30:27the shift in robotics that happened
30:30I don't know in the late 80s Early 90s
30:31at least with respect to motor control
30:34where the intelligence like in hexapods
30:38rather than having a lot of intelligence
30:41in the central control the intelligence
30:43was moved out to the legs and and and
30:47this was this was then picked up by a
30:49bunch of toy companies I think I can't
30:52remember the names of the toy companies
30:54right now and they're not paying me to
30:56advertise and so I won't but but I think
31:00people probably are aware of this
31:03because these kinds of results been out
31:04for about 20 or more years but when you
31:08win the when the the shift happened
31:11where the intelligence was decentralized
31:13out to the legs then for a wide class of
31:19robots what we saw was a much much more
31:22robust mode motor control so that's a
31:26and what we're suggesting here is we're
31:29moving the intelligence out to chess the
31:31pieces right there's a there's a sort of
31:34little mini intelligence in in each pawn
31:37or driving each pawn there's a little
31:39mini intelligence driving the Queen or
31:42the bishop for the knight and and
31:44consensus is what glues them together to
31:47act as a single player and that's that's
31:49kind of the the approach here and notice
31:56that if if Y player cannot provide such
31:58a proof then black.there is justified in
32:00assuming that y players really a
32:02singular agent right everything else is
32:04just a sock puppet and that has that has
32:07import for how we engage into these on
32:13the Internet
32:14which I think you know there's been a
32:16lot of discussion debate around that now
32:18no Otis that in the art chain setting as
32:22opposed to other settings we can nest
32:25these sub communities so it's perfectly
32:28reasonable that a particular set of
32:32clients that are only interested in a
32:34single pawn live in their own namespace
32:37and and then because of the way that our
32:40art chain works you can aggregate these
32:42up into a larger or more inclusive
32:44namespace for the consensus across the
32:48total set of client proposals that then
32:51roll up into a single loop and then you
32:56know there's some other there's some
32:58other niceties about this you can think
33:01about piece removal in playing the game
33:04as one of two different outcomes when a
33:09piece is removed it can either be a
33:11validator as cashing out or unbonded for
33:13the protocol or it can be a network
33:15failure and in fact if a pawn makes it
33:18all the way if another pawn makes it all
33:21the way to the end of the board and
33:22brings another piece back then that
33:25could be seen as as a validator cash out
33:31oh sorry
33:32as the network failure and otherwise you
33:35know the piece never comes back on the
33:36on the board that was validator cash out
33:38so there's some some niceties about the
33:40fit between the the tests approach and
33:44some of the features that of Casper that
33:48we're trying to test additionally when
33:51we look at other games for example the
33:53game of go there are there there there's
33:55more to this story specifically when we
34:00in the game of go unlike the game of
34:03chess new pieces are being placed on the
34:05board and that corresponds to validator
34:08bonding so testing the validator bonding
34:11at least in this way of factoring the
34:13architecture so that that has that has
34:16some some additional benefit when when
34:20we think about this just from the point
34:22of view of testing Casper
34:26one interesting point is that this
34:30doesn't do look ahead right so what
34:34white player as an organizational player
34:37that there's nothing in this approach
34:39that does any look ahead in terms of the
34:40gameplay so it's quite likely to be a
34:43lousy chess player but I'm interested to
34:46see how bad it is and I would leave that
34:49to future iterations
34:53also note that disparate stake
34:57distributions will lead to different
34:59player strategies so if the Queen is
35:02heavily staked right then Queen moves
35:05and end up winning more frequently
35:08unless there's you know some kind of
35:11collaboration amongst you know say the
35:14Knights or the the Knights and the
35:16bishops or things like that so in in in
35:21that case you'll get a different profile
35:25then let's say if the rooks were more
35:27heavily staked where the King was more
35:28heavily state so the distribution of
35:31stakes will result in different
35:34gameplays against the opponent and I I
35:37just kind of you know I mean one of the
35:39things that's that's fun about these
35:41kinds of tests is is you know I'm again
35:46I'm trying to game a fire and make it
35:47fun to to to look at lots of variations
35:51because if we do look at lots of
35:54variations that we're gonna end up test
35:55and getting more coverage of the Kaspar
35:58algorithm and then kind of the the the
36:04important point with respect to the
36:07implications of consensus for society is
36:12that we can begin to contemplate
36:13fruitful interaction between singular
36:15agency like a bear
36:17searching for honey such as produced by
36:21an organizational agency like like a
36:23beehive and and you know I mean
36:27obviously nature is full of this nature
36:30nature is full of predators Co evolving
36:34with herd like prey and
36:40and it's interesting to look at Homo
36:42sapiens as kind of balanced between you
36:44know being a social animal but also
36:46working with singular agency as well and
36:50in fact that sort of picks out human
36:53beings and Homo Sapien says as one of
36:55the one of the more intriguing species
36:57because it because it is dual in this
37:00way you know it's it's both singular and
37:04and social and and kind of what I'm what
37:10I think is going to be very important is
37:13that this class of of consensus this
37:16algorithms is going to make it possible
37:18for a lot of the creatures in the
37:21Internet to to take advantage of both
37:23modes of agency so that's that's kind of
37:27the the rough outline of the idea or the
37:30proposal I'll stop there and open the
37:34floor for discussion where people able
37:39to follow that yeah that's that's really
37:43cool Greg I have a question at least
37:47initially when you said that say you
37:51three clients that are responsible for
37:52proposing moves possible for one piece
37:56are all responsible for submitting that
37:59move for consensus we need to be a
38:02transaction so how is one move chosen so
38:07since they're competing right only one
38:11of them gets to be chosen right and the
38:13others the others are effectively bad
38:16debts right so over time the old group
38:20has to converge to one of those moves
38:23because that's the the block size is
38:24exactly one so there's only there's only
38:27one transaction amongst all the
38:29transactions proposed for the pawn which
38:33is also being thrown into the pot
38:35against you know all the transactions
38:36proposed for a night all the
38:38transactions proposed for the Queen etc
38:40right so they're all all the validators
38:43are engaged in the cast for betting
38:46process to select exactly one of all
38:49those proposals right
38:53but potentially you could like say if I
38:54want an outfit the say like the consent
38:58this protocol with a certain I don't
39:01know logic that specified or constrained
39:04moves to an optimal move right because
39:07in chess we want to win then I could for
39:10I mean like possibly constrain whatever
39:13transactions or proposals that I was
39:15getting to fit within the constraints
39:17with which we define our optimal move
39:20yeah yes and pointed back to your your
39:25foreshadowing the next level of
39:28development that I've wanted to kind of
39:30keep in my back pocket which is when we
39:33when we utilize our chains urgent
39:36[Music]
39:37Kaspar where we have formulate that
39:40describe what goes in the block you can
39:44now use the formula to describe you know
39:47information about a game play strategy
39:50round I don't want to I don't want to
39:54open that door yet I'm trying to keep
39:57their curtain closed just for a minute
40:01and this is exactly what I'm hoping for
40:04people to pick up on so you know you're
40:09on the money as usual very good you're
40:13blowing my mind again Greg but I'm still
40:17not completely clear on who's doing the
40:21betting oh so so the betting is done by
40:25the validators let me go back to screen
40:27sure and know and they want to win the
40:32game validators are completely unaware
40:37of the game of chess they have no clue
40:40that chess is being played
40:44only and in fact there's no agent here
40:49that sees that there's a game of chess
40:51being played right so in this schema
40:53they they reach consensus on a random
40:57move essentially yes it's not actually
41:04random because the validator days are
41:06staked right right I mean that doesn't
41:11make it any less cool that's still like
41:12an I think an awesome way to think about
41:14it right right exactly and so what you
41:17can do is then you you you can you can
41:20make it less random and and how you make
41:23it less random is actually quite
41:25intriguing higher in the game that's
41:37tweaking the game and that's what I said
41:39the the the the the the staking
41:41distributions changes the player profile
41:44but that's not sufficient right it's not
41:47sufficient and like you know pulling
41:49your queen out first again certain kinds
41:51of opponents is not going to be a good
41:54idea all right so so there's there's
41:59more to this than just the you know
42:07favoring the move of one kind of piece
42:11versus another there has to be a way to
42:14encode more of the intelligence of the
42:18whole and and that's kind of the next
42:21step but I just want to get people used
42:23to the idea that there is no entity here
42:26that has a global view of the game right
42:33so it's like let me know know I'm kind
42:36of an intuitive level it's like the
42:37pieces are playing the game rather than
42:39the person behind them that's correct
42:41the pieces are playing the game and
42:43that's what I was trying to say that
42:45this is remarkably like the move in
42:49robotics so the legs don't know that
42:53there's an ant
42:55right the legs don't know that there's a
42:58there's a goal seeking behavior by the
43:01ants the legs are just trying to be legs
43:03but they're they're smart enough to be
43:05Minh you know fairly good legs and that
43:10that makes that kind of decentralized
43:13intelligence at least you know for it
43:16was a major shift in robotics and and
43:19actually I was talking to two glad about
43:21this and he was asking for references
43:23and where I found out about this was in
43:26Kevin Kelly's book out of control which
43:29I read in the early 90s and I heartily
43:33recommend it to anyone who's interested
43:34in decentralization Kevin Kelly was way
43:38way way ahead of the curve on this and
43:41so I I really think people you know I I
43:44don't mind you know promoting promoting
43:48that book because it's worth reading it
43:51really talks about the shift that the
43:56blockchain I think is is is very much a
43:58part of we use decentralization push the
44:02intelligence out to the leaves more
44:04autonomy and and effectively you know
44:10it's a funny relationship to trust
44:13because what your end up doing is you're
44:14trusting your peers rather than then
44:18trusting in some overarching control
44:20that governs all the peers yeah that's
44:25that's so if we could define in row Lang
44:36what it means to win a game of chess
44:40could we use that for staking
44:53III didn't hear you
44:56Jim you cut out can you can you repeat
45:00sorry I'm sorry huh
45:02yeah so if you could write in row Lang
45:06define what it means to win a game of
45:09chess could that be applied to stinking
45:16yeah I mean you can you can do anything
45:19you want you know in a Turing complete
45:23language you can do anything you want
45:25the important the important point here
45:27is to not bend pass per out of shape
45:31right that's that's really that that's
45:34the critical thing is like you don't
45:36want the validators to know that they're
45:37playing a game of chess right you want
45:40the validators to only be concerned with
45:42consensus so all of the intelligence
45:45about playing chess has to be has to be
45:49pushed out to the leaves although as I
45:54said just by just by having this one
46:00constraint which is the block size is
46:03one that forces the the alternating move
46:07thing so there's there's lots of
46:09different degrees of freedom here that
46:11one can tweak in order to make this
46:14organizational entity more intelligent
46:17and a better player but but I again kind
46:24of the important point here is that it's
46:28not that that that we as we try to make
46:32this a better chess player we don't lose
46:34the fact that the validators are just
46:37just giving consensus so so what one
46:43thing the one thing you could do for
46:45example is that you could make the pawn
46:48or the clients that move the pawn around
46:51more aware of the whole game right so
46:55for example if there were if there was
46:57human agency behind each of these
46:59clients as opposed to AI then
47:02the humans could all be looking at the
47:04chessboard right and so you know now
47:07they know they know that they only can
47:10move their pawn and so if they think
47:13that their pawns move is a good one
47:15they'll propose if they don't they'll be
47:18quiet right and and and that's that's
47:23kind of how how this might change if you
47:26had greater intelligence at the client
47:29side but but that's again I'm just
47:35trying to be suggested here I think that
47:39the the most important point is
47:42beginning to understand that that
47:49organizational intelligence can be can
47:54emerge out of very very restricted moves
47:57very very small things I mean one of the
48:01one of the places where this landed
48:03absolutely I mean made it crystal clear
48:05in my body that this was the case was
48:08when I went and studied West African
48:10drumming in Africa the any one part in a
48:14sub R piece is easy I mean it's very
48:18very simple
48:20the village idiot like myself can play
48:22it and yet when you put all the pieces
48:24together the complexity of the
48:27conversation is what makes the song
48:29satisfying and compelling and and and
48:34that's that's what I'm trying to suggest
48:37is going on here but it's also the fact
48:41that any participant is getting feedback
48:44from the whole that that makes this that
48:49makes this begin to work and that's kind
48:51of like what I was suggesting before
48:52that the the the players that are behind
48:56a particular pawn if they had a certain
48:58kind of intelligence or agency they
49:00could be looking at the whole board the
49:03players who are playing sub are in a sub
49:06R orchestra they can hear the whole song
49:11and they know how a subtle variation in
49:15their their very sim
49:16hard will either you know bring out you
49:20know a conversation or subdue a
49:23conversation or provide provide some
49:25kind of nuance to to the whole to the
49:28whole piece so that's a it's it's a
49:34different kind of thinking about
49:36organizational intelligence um that I
49:39believe you know goes back to the
49:41origins of our species that we've always
49:43understood this and we we kind of keep
49:46forgetting it gives am gives a great
49:50talk it's called the four pillars of
49:53decentralisation we're essentially
49:56argues as you know whether the current
49:58hierarchy that exists between or that
50:01exists in our society kind of the
50:03patriarchal like you know political
50:05landscape that is you know that's how
50:08humans organize ourselves now and it's
50:10essentially that in the early you know
50:14in the early part of human mankind that
50:19we existed in small kind of
50:20decentralized nomadic tribes and such
50:23and you know in that time everything was
50:26mostly decentralized so it was pretty
50:28easy to get consensus on an entire group
50:30but you know as as kind of our
50:33population grew so if we fast-forward to
50:36Rome and the size of Rome was it became
50:40technologically essentially impossible
50:43to get consensus on such a large group
50:45of people we just didn't have the tech
50:46for it so a hierarchy was essentially
50:50the only way that we knew how to scale
50:52our societies that is like you only
50:53answer to one person above you and you
50:55can't you don't pay attention to what
50:57essentially outside what's outside of
51:00your scope which kind of he argues you
51:03know let by the way for for certain
51:06individuals to claim that the hierarchy
51:08was broken and then for people to vote
51:10to replace the hierarchy and hence you
51:14know we see essentially that you know
51:16any modern election is always about
51:18change right and it's because everyone's
51:21it's it's it's a time old tradition to
51:25say that the hierarchy is broken but I
51:28think it's interesting now that
51:30were finally at a point kind of where we
51:32have the technology that is enabling or
51:35that would have first enabled the same
51:38decentralized nature that existed in the
51:41early course of mankind so it's called
51:43the four pillars of decentralization
51:44it's pretty good talk
51:46cool I'll definitely check that out
51:48absolutely and I think you know bringing
51:51it all the way back round two to how I
51:53began the conversation whether I'm
51:57talking about what's there freaking
51:59drumming or I'm talking about two-person
52:01games
52:02the thing that I'm trying to convey to
52:05get across is this is critical
52:10understanding that the dialogue between
52:15individual and organization between the
52:18one and the many can be many times more
52:23satisfying and enjoyable it doesn't have
52:26to be this mind-numbing ly frustrating
52:30or spirit-crushing ly frustrating
52:32experience an organization can be
52:35responsive and and and playful and and
52:40engaging right and we know this like
52:44that's why people go to see music right
52:47when we go to see a rock band or we go
52:51to see a jazz band and we see the
52:52players having fun and out of that fun
52:55and joy with you know comes this
52:59fantastic music right you know that's
53:02what we want to see we see that there's
53:04there's there's a right way to be an
53:09organizational intelligence and to
53:11participate in orientational
53:12intelligence whereby right I mean joyful
53:15and productive and and and and having
53:18having a purpose and and and and the
53:24sort of to finally put the final point
53:27on this this is the kind of
53:30organizational intelligence that we have
53:32to have to face the oncoming changes so
53:36if we want to be able to respond
53:39effectively to climate change we have to
53:43have a
53:43fundamental shift in the nature and
53:46quality of our organizational
53:48intelligence and that's that's why the
53:52consensus work is so salient and so
53:56necessary at this point and and so again
54:01you know I try to try to bring bring
54:04this all together I'm just talking about
54:06a test framework for Casper that we can
54:10literally code up together right but
54:13once we have that test framework I
54:16believe that people's eyes will be open
54:19as to the implications of what
54:21decentralization and and the blockchain
54:24technology means in terms of the
54:27existing communications and governance
54:31infrastructure I've got a quick question
54:36yeah you mentioned that the black player
54:40can ask for a proof of consensus which
54:44is going to be a sequence of justified
54:47messages is it correct that the sequence
54:50of justified messages would constitute
54:52an explanation of how the one moved was
54:54chosen out of the many moves that's
54:58correct
54:59and we've heard okay it's not a single
55:03sequence it's a set of sequences sure
55:05and we kind of deferred discussion in
55:08this document of how the choosing
55:10happens that's correct because because
55:16that discussion has been ongoing for the
55:18last two years it effectively it's a
55:20form of it all right so so each each
55:25block is is proposed and the the
55:31validators participate in a betting
55:34cycle on the proposals for the blocks
55:37then and as as a validator sees new new
55:42proposals those proposals have to be
55:44justified
55:44[Music]
55:48and and there are conditions on when
55:54proposals are kind of in good faith for
55:57versus when proposals are clearly in bad
55:59faith effectively those proposals which
56:02have the same justification but propose
56:04two different the two different moves
56:09and in this setting it would be two
56:11different moves same justification
56:12that's equivocation and the equivocation
56:15is is always considered in bad faith and
56:17is always slashed as then the
56:21requirement is that the validators must
56:25gossip amongst each other which then
56:29basically causes a reveal on
56:31equivocation and once you have a reveal
56:34and equivocation that's when you you can
56:36you can be assured that slashing law
56:39will take place because it's it's in
56:41everyone's economic interest to slash
56:43someone who is equivocating or a
56:45validator that's equivocating so that's
56:47the rough sketch of how of how the
56:50choice happens got it thank you sure
56:56cool any other questions comments
57:01thoughts Thanks oh well thank you no
57:08it's it's great to be able to be able to
57:10chat about this I really well you know
57:29if we're going to be inspired by anyone
57:32I'm taking inspiration from Martin
57:34Luther King is it you know that's that's
57:36one of the best sources of inspiration
57:38one could possibly have so cool thank
57:41you
57:42well I'm fully intend to put a bounty
57:47program around this of course the first
57:51thing I'll do to follow governance is to
57:53make sure that I make a proposal to the
57:56board and get feedback from the board
57:59and assuming that the board is cooled
58:01that I would like to put together a
58:03bounty program that goes out on the
58:06heels of having committed the Casper
58:09contracts to github to begin the
58:12development of the the test framework
58:14and that's where I really really want to
58:16have the community as much as possible
58:19involved in the development of that test
58:21framework which is why I think a bounty
58:24program is like a first step in doing
58:27that so that that's kind of all I got
58:34for this week I'm gonna jump off because
58:38I've it's already 9 o'clock here but I
58:42you know please you guys stay on if
58:44there's more to talk about
58:53ciao for now
58:54take a rest