 Thank you very much for having me. It was a very last minute, but I made it and it's great to be here So I'm presenting a vision that is being very much materialized And be released shortly and it's a platform for collaborative scientific discovery So instead of it's it's quite a generic solution for a lot of scientific problems And it's got a lot of access along which those problems are solved so instead of starting with the list of problems which have been listed on many occasions during this conference already I'll start with a vision and As a scientist Or as a thinker as a person I want to contribute to the body of knowledge in any way I can and anywhere I can and by anywhere I mean not like on the beach, but Rather any way in the body of knowledge or whatever I can see solution That's the kind of generous approach To I think participating in research or knowledge production, and if I see a solution I you know ideally I feel generous I want to offer it and then people can go away and use it and I'll be happy for that but current system prohibits me from doing that because I'll be screwed over and I will never get any benefit from that And if I'll just be generous I will not earn any living so to be Generous I need to trust people that I work with or the people that I offer solution To and I need to trust the knowledge that I build upon so The knowledge you know the solutions that I provide They need to come from a solid knowledge base Which is not false. Otherwise the solutions will not work and I will have wasted my time and people's time And also in this situation. I want to be earning some Revenue and be rewarded for my activity But still going back to discussing the problems in science a little bit I want to point out why do we have these problems and What is science I define science more than as a method rather than a subject So if you're looking down the microscope, I don't think you do in science necessarily It's only if you're applying the scientific method to it then you're doing science You're validating you making some assumptions you're coming up with hypotheses and then you testing them and That that is science not a specific subject so to Scientific method is quite hard. There's a logical structure and observational structure. There's measuring to tools which need to be sensitive and valid So to maintain scientific method we build hierarchy of people who we trust To be very logical and to know what the good measurement is. That's how historically Arisen and they maintain this method by Controlling the people that they work with and validating them. That's what the hierarchy of scientific research that we have But unfortunately It all evolved into gaming the system. So all these people who are high in hierarchy the actors bottlenecks of Flow of knowledge and if you're in a typical lab, there will be postdocs and PhD students And the BI will be on top and you can only communicate your knowledge through publications Which the PI will control whether you can do it or not and the funding Is also controlled by the PIs for example in the UK. You cannot apply for a funding if you do not have a tiny position And then in terms of collaboration with other people if you belong to two labs and you do your PI's do not Get along you cannot possibly collaborate So you can see how this hierarchical structure instills the bottlenecks of the knowledge flow and Stops collaboration which I think prevents Faster progress And they all do this of people in hierarchy do it or we all try to do is to leverage our position and reputation for our own benefit rather than for the advancement of knowledge and progress and a lot of our Production contribution is evaluated on our name on our reputation Which is an indirect measure of knowledge so This is the old system or the current system that we're talking about which is on your right And it's a person-centric system and a lot have been said about standing on shoulders of giants I'm not sure about that but definitely what we're doing is we're standing on each other's shoulders And that's how the current system is structured. So we have a very highly structured people organization Which produces a lot of output? that via the publishing Infrastructure goes out into this poorly organized body and here I represent the Hierarchy or research establishment in the center In a pyramid and then the output goes out into this cloud of papers which are Very tenuously linked to each other So what I'm proposing is to reverse this structure and To build a knowledge centering system where the knowledge is highly structured and it's at the center of this Organization and then you have the protocols Which are secured by the technology that we're discussing the blockchain. So we're transferring the Guardianship of the scientific method from hierarchy to the protocols and then we no longer need this hierarchy and then people can be in this cloud instead of the knowledge and I don't know about you, but this reminds me of some sort of visualizations of entropy and entropy is related to How many possible combinations each System can have and in the person centering combination System the knowledge has lots of permutations. So because it's not very highly structured. It's like a gas So there's a lot of uncertainty about the knowledge and that's what we're experiencing with poor reproducibility and stuff whereas In in the knowledge centric system the knowledge is much more structured. So it's has a lot less uncertainty, but The people structure has lots of permutations. So you have lots of possibility for collaboration I Do not have a formal proof to this, but I'm working on it So here it is in the system represented. You have certain sees we building the system along Around the knowledge graph, which is tightly kind of structured system of knowledge and you have protocols that are shared and enforced by the network And you have a value distribution incentivization that's built around the graph and The freedoms that you get from the system is the project participation. So each person from this cloud can via different protocols can Make a contribution to any point in the graph There is no reputation. So the whole system is anonymous. So it's impossible to leverage your position like why the name or anything like that So it provides more freedom and more opportunity for collaboration this is the Basic structure of the system. So it's a fully decentralized peer to be a network And you have machines that run what we call a peer and you have users represented here in as human Human figures they can either own the machine that runs a peer or they can connect to nearby peer and kind of the stack is We use hyperledge fabric for blockchain and network IPFS for storing heavy kind of data Loaded files and raw data sets Then we have an IPI and UI that communicate to each other and to the system components To allow use interaction. So we We try to make the system as adaptable as possible because as we can see from many presentations People have different use cases and they design different protocols before that was genomic protocols So you have specific needs like protected data structures and Limited access to it and different validation protocols that you need to apply to that data So What our system allows which is quite different from previous presentations is You have custom protocol specifications. You can write your own protocol. How you validate stuff How you analyze stuff and so on and you can use any arbitrary language. So you're not limited to a Smart contracting of the fabric platform or whatever you choose And you can run the code regardless of the environmental operating system So this is what we call modules Modules are those bits where you kind of specify how your system how you want to run the system And you have as I said, you have validation computation analysis and graphing functions and any user can choose To interact with the system with whatever this would be like a selection of different modules on the system And you can plug in any analysis of computational software Into this core system with a blockchain so that any Open software or any other software in for scientific Research and computation that already exists actually can be used on the platform You have then customization protocols and custom graphs. So you can do graphs manually or can you can design Functions that draw the graph for you for example phylogenetic graphs or I don't know molecule chemical molecule relation graphs You have customizations like your draft. So basically you can adapt the user interface to your specific Use case so that it's really useful for you And you can play with custom token value distribution Which we keeping open at the moment it wouldn't work in a big system But because we're designing the system and we don't know what the protocols would be best suited to scientific research in an open environment. I think that's important to keep open so to give an example the When you have a contribution to submit to the network you define you specify. What's the contents on the node are so you have Datasets some sort of logic claims and the logical relations between those and You also specify the position of the node that you prove the contribution that they're proposing in the graph so I specify the contents and the nodes connections and When it will be accepted you You know your authorship will be permanently linked to that node So you both know what you have contributed, but also more importantly you would know Where a contribution lies in the context of the rest of the knowledge and it's quite fixed So if you think for example when you buy a real estate property the land registry records not only what your house is And maybe it's like less interesting even But you know you have a balcony and whatever three bedrooms, but it also tells you what the address is and it's very important just if you Only specify, you know the contents of the house you would not be able to find it and you're not really protected So when you submit the node it has to go to validation and the two types of validation machine validation and Human validation machine validation Wherever we can use it for to validate analysis or graphing functions I mean outcomes of the graphing functions and the human validation is pretty much a peer review which we have adapted and We have a kind of game theoretical protocol for peer review where author nominates a bounty and then the peers or reviews are selected from a fraction random fraction pool, so you have minimized the potential for Kind of rigging the system by your friends reviewing your positively for example Reviews have to pay stakes Before they doing review and they have opportunity Well, there's probability of them losing them if they do not perform well if they do not produce good work around time or Even for example in cases say for example, we have six reviews and four reviews voted yes and two voted no Of valid and valid then the minority will lose their stake and the majority will receive it so that everyone Maintains an interest in actually doing work in this validation protocol and then afterwards anything Within the graph can be challenged so Peer of you can be challenged After this protocol and all the nodes and their connections can be challenged and remodeled and there is a kind of curation market slash Game theory protocol or behavioral economics behind it So and that provides accountability especially in the case of peer review because usually what happens now Is peer review is done and that's it you forget about it and nobody talks about it anymore so Lots of peer review is kind of an unfair, but there is no way to appeal about it and Once the validation has come through and if in case of acceptance the network then reaches the consensus on the node Content and its position on the graph and consensus on the graph itself. So everybody on the network Has exactly the same information about current state of knowledge This is kind of the basic Workflow of something that would be similar to a paper Submission thing so How the community can engage with it with this platform? So say for example, you have historic literature put it into the platform and then Small or slightly larger community can resolve it into a more efficient graph Then they can put proposals on top of it represented by P's here and then Ideally that would go to a funding body. So it's almost like making review and getting your implications from the historical literature and then you go to a funding body say well this proposal seemed good Could you allocate potential funding to this and the funding body can allocate? Money where which in case could be tokenized or transferred in tokens They can allocate the funding by instead of allocating to one lab It would allocate it to a body of knowledge that can be produced in the future and anyone within That domain can get the share of that funding So you never bet only on one horse there and then in the end You know after some time passes the community the funding is allocated the community builds extension to this graph and You know new data produce any nodes produced and then some milestones or breakthroughs reached like discovery or a molecule for example and the funding arrives at that node at that point in the graph and it trickles down to the rest of the Discovery chain that participated in the production of this Graph so in that way you no longer need to hire what are you doing? In isolation Until you reach this breakthrough point you can contribute at any point of this graph and let somebody else Reach the major breakthrough but it would still be rewarded for your contribution in this chain of discovery and a slightly more So that that's kind of a general View of how the platform could function In the same way like we're writing research papers, but on more granular level But here's a more specific example where things are a lot more specified So this is a this is a project that we work with with pathogen surveillance network and these graphs are representation of Ebola outbreak and each node represents a sampling at clinical centers of different pathogens of Ebola virus and then it's Assembled in these phylogenetic graphs, or you can you know specify where it is Geographically or you can have also time series and see how the Spread of the disease happened so what I'm trying to communicate here is On this network with this software that can be plugged into the core you can have distributed computing so say for example somebody sequence genome in Africa and The genome sequence you put on the network There are some machines that can come in and analyze it for a genetic variants to see Which clones are which and whereabouts they are in the geographical locations? You have automatic graphing functions, so you no longer doing it manually like in the historical literature example But The network what network allows you to do is to reach consent consensus of methods. So say for example, you all know that the genomic Sequencing data is valid and it doesn't have huge differences that would allow artifacts And you have again a consensus of graph data So everybody knows that the graph that we're working on is this one and not some other graph that you generated locally And you can use these graphs for custom value assignment economic incentivization So say for example, you found that the new particularly infectious clone is emerging is in Guinea You can incentivize that whole part of the graph Or you can increase the incentivization and hopefully a lot more sequencing Contributions will come from that area So we've got a functional release that's coming out in February 2019 where we'll have two modules this kind of paper module and then pathogen surveillance module then will experiment With the network and see how we can play with real situation and see what the outcomes will be and We'll invite anyone to engage with this by you can come and write your own modules or experiment within your communities in terms of research Both on science or on behavioral or economical models. Thank you very much So We are quite well on time maybe we can have two questions and then we have a coffee break and for half an hour So I I can ask a question. So are you at which stage are you currently with the like? Implementation of this is it Well, we developing it And it's the functional version will come out in February. So the stages I mean we've done quite a lot of kind of back-end and network establishment and communication between IPFS fabric and design some ways that and tested some ways that allow You know programming language agnostic interaction of different components I Didn't understand where the content will come from are you going to start by new findings with new findings or? Well, you probably start with historical literature. I mean nobody starts with new findings any papers has introduction so You always have to put your findings into a content. I mean you can probably if you want start with new findings historical publications I meant Sorry The access to historical publications to publications that have been published before Yeah, scientific findings. So I didn't get how you will bring them in into your system Manually or by some sort of scraper It's I mean say for example you you have something to contribute and you want to put some historic literature that you take five papers you put them into the System if you want to resolve them into more granular graph if you don't want You just leave them as they are Thanks. I wanted to ask in your peer review section. You mentioned Things like all the bounties and you also mentioned like Random fraction polls. I was just wondering how you're going to use that Well, when you have a submission you kind of get a template of what you're submitting so data sets and whatever and then you specify how you want your peer review being done and then you Nominate a bounty. So there's a native token on the platform. So I say I don't tend tokens which will be equally divided between peer reviewers and then to select the peer reviewers a call goes out to a random fraction So the you can never predict who is who is actually gonna receive a call And then let's say you specify that he wants done a five peer of yours. So they come in The first five to self-recruit Will be in your review pool. That's how we're doing it at the moment, but maybe there's a slightly better Model of doing this. I don't know. So I love the idea and organizing knowledge. That's pretty great How to how are you going to make sure do you have any plan for it that if people put in Their knowledge their pieces of knowledge, whatever it is going to be data or Publications in this sense also historical that they're not going to be Overlapping so that in the end we really make sure that we build up on the knowledge and not just built kind of parallel chunks So they're not connected Well, the community says there's like what you call a curation market and I call network cohesion protocol So they you know anyone any user or all the users are incentivized to make the graph more Coherent so resolve redundancies or where you need maybe some nodes are too chunky you need to separate them and create two out of one and also, you know Maintain the right number of connections and make sure that they complete Does that answer? Who are going to be those curators that sense? Anyone the the users on a platform are anonymous and anyone can participate In that process, so it's not there are no there's no reputation. There are no experts They're only economic incentives for right behaviors and their penalties for wrong behaviors on the platform so if you choose to do something you Have a chance of earning something but also you have a risk of losing it So it's up to you if you feel an expert enough to go for it Thank you Will does this or will it have an AI system associated with it? It can have yet. So you can have actors could be a eyes I'm not specifically building a eyes Here but the AI's can be plugged in and into here. Yes, and would that be open source? Would the algorithms be open source? Well, the platform will be all open source You know the core and all the modules will be open source And you can build your own modules and we're building like a dummy module which can easily be Taken and build up into whatever you need. Thank you. Thank you very much