Projects
![]()
No projects found
Investors
New
DAO Treasuries
DATs
Lend / Borrow APRs
Chains
Categories
Spaces
Interoperability in DeSci: Connecting projects, data, and workflows
Recorded: Jan. 29, 2025 Duration: 1:06:51Space Recording
Full Transcription
Thank you. Thank you. All right, I'm super excited to dive into this episode of the design mic.
This is Erin McGinnis behind the account.
And this week we'll have a bunch of different awesome projects on sharing a little bit about
what they're each building and then diving into how might we create more flows between
each of these different projects, whether it's on the
data side, knowledge sharing, creating meaningful insights from it, adding verification, connecting
some of these different communities together, and really pushing DSI, open science, scientific discovery, all of it forward. I've sent out a couple of speaker invites,
but I don't know if those have gone through.
So if you're able to request to speak,
then maybe I can improve that way to kick this conversation off.
then maybe I can approve that way
to kick this conversation off.
In the meantime, there are a ton of other desized spaces happening today,
and so definitely check these out as well.
So many great conversations going, which I think just tie nicely into this overarching theme of DSi's moving forward and adding this different interoperability.
So many great conversations going,
A lot of these spaces are also collaborations across a handful of different organizations.
So that's super exciting to see more and more of those kind of joint conversations happening now in more public forums and being able to invite other people to participate as well.
Okay, I just got a message that at least one speaker invite wasn't able to be accepted.
speaker invite wasn't able to be accepted. Let's see if we can get some people up here,
because I know some of you are down listening in and should be up here on stage with me right now.
Cool. We got one person. Hey there, welcome.
Cool, we got one person. Hey there, welcome.
I think we should have more speakers.
Yes. Yeah.
Cool.
I just got off of a meeting right before this with
Desai Labs actually talking on this topic as well
of interoperability between CausalityNet and Desai Labs. So really excited to continue kind
of that vibe of conversation with everything you're building over at Index Network. Levy welcome um maybe to kick things off would love for each of you to give a brief overview of
what you're building and maybe some of the inspiration behind uh why you're building this as
well. Levy would you like to jump in, sir?
Yeah, sure.
I'll go first.
Yeah.
Thank you, everyone.
And thank you, Aaron, for hosting.
Yeah.
My name is Levy.
I'm the founder of Co-op Hive.
We're building a way for any agent, human or machine, to make programmatic commitments to each other.
And what that means is really enabling a way
for agents to make an agreement over any type of objects
or real world event,
or specifically in our initial use cases,
compute and storage and bandwidth with other agents.
So like, for example, we're like making these marketplace primitives that enable you to
create a marketplace for compute or storage or bandwidth or data or NGO real world assets
started from these basic building blocks.
And one of our core focuses is on scientific computing.
And as a sub point of that, it's reproducible scientific computing,
right? Because we have this massive replication crisis in science right now, and we need a
way to reproduce the outputs of jobs in whatever way that might, you know, that's necessary,
basically, in order to ensure people that like the results are valid and accurate. I
say that's, that's, that's part of what we're building.
Let me continue then.
Again, thank you Erin organizing this.
Pretty excited.
I am Sheref from Innext Network.
We are building a protocol for discovery.
So what do we mean by protocol for discovery?
Let's say we have telephone.
Telephone works when you know the number, right?
But what if you don't know the number?
So we need a protocol for finding things.
Today we use search engines, dating apps,
or for scientific discovery.
Like we don't have a protocol for discovery.
So super quick.
Like it's basically a distributed semantic indexing layer
for the internet.
So like we are building the first decentralized semantic index.
It can be used for scientific discovery,
for collaboration, for agents,
enable them to collaborate.
Yeah, I think that's the shortest summary.
Thanks.
Loving all of these different overlaps,
especially just kind of trying to address
the bigger topic of the reproducibility crisis.
So top of mind, I think, for all of us, and then building out some of
these different networks to connect everything together in meaningful ways. I just got a message
that one of the other speakers isn't able to make it right now. He's from Coordination Network,
which is working with a lot of different groups in the space as well to kind
of bridge some of those different connections, especially to insights from different data,
whether that's audio, text, other types of data as well, and create kind of more sense between all of those different materials.
So kind of another puzzle piece in this whole conversation.
For anyone who is listening in, as you're hearing different comments,
feel free to post any questions or ideas you might have down in the replies or request the
mic and we can bring you up on stage. Some of our intentions for this space are to, one, kind of
lay out what are some of these different technologies that exist just so we can all
have more awareness of them and then start exploring okay how might one of
these initiatives or projects or technologies connect to what someone else is working on and
then also filling in gaps kind of across the broader scientific data discovery conversation and those gaps is really where I think everyone can come in and
either identifying those or kind of pointing out hey I see this cool way of how these two different
technologies might be able to connect together. Awesome. Causality Network would love for you, Barrett, to give an overview of what
we're building at Causality.
Hello, hello, hello. Thank you for the invites. Sorry, I had some technical issues. I hope
everyone can hear me. Give me some thumbs up or hearts yep cool um yeah so the
i'll start super broad um the area that we are operating in uh we're referring to as truth tech
and with the you know upcoming ai scenarios that we're already seeing in social media for example
you are not going to be able to trust Did his audio cut out for other people as well or just me?
Also cut out for me.
I thought it was just me.
Oh no.
Now we can hear you.
We can hear you now.
I think I just had to call it out.
Okay. Okay. Yeah. now we can hear you now we can hear you i think i just had to call it out okay
okay yeah i'll i'll i'll just do the tldr protocol it runs on hardware so any kind of scientific equipment and cryptographically signs the data at source so you know the data is real um and
overall this is we see the kind of scientific process moving away from trusting.
Did he get rugged again?
Yes.
Oh no.
No, now you're back.
Wait, okay, Erin, you are the trigger for reviving
Cazamarin Network's Twitter voice.
That's so funny.
Each time I say something, then we can hear you just say, oh, no.
I've shared something, so maybe just look at the picture.
It's a nice picture.
And then, yeah, I'll mute myself.
Cool.
Where did this get shared?
I want to pin it up above. Awesome.
Yeah, so kind of tying in or filling in some of those different dots at causality,
really focused on that raw data generation layer,
and then see that being able to funnel into some of these other platforms that might either be focused on the publication of that data, generating insights from that data, and really kind of building out this initial truth layer for it.
All of these other things to operate kind of down the line from.
kind of down the line from. And we also see this being able to help unlock both DSI projects,
have more legitimacy in the eyes of traditional organizations or just kind of more legacy players
who might be a little resistant to newcomers coming in, as well as so giving that legitimacy to dsci orgs as well as individual scientists or
unlocking the scientist creator economy is kind of how we position that so that's some background
on causality network see a couple people down below from dsci. I know before I was having some issue creating those invites up,
so if anyone from there is able to request to come up, would love to have you
share from Labs' perspective as well just to lay out some of these different puzzle pieces.
different puzzle pieces. I guess so far, Sarah and Levy would love to, and Barrett as well,
would love to kind of talk high level about the reproducibility crisis and what different
opportunities or things might need to be implemented to address that across the board.
to address that across the board.
Maybe Levy, you kind of mentioned that term first,
if you wanna kick off any thoughts first.
Yeah, I mean, so I think that there,
and I guess this is kind of like a larger retrospective
of the last two or three years in V-Sci,
but there's been like a lot of us who are super passionate about the space and we all want to work together.
But there weren't, it was clear that like in our individual or even collective,
in our individual or even collective smaller collective group kind of network nodes that we
were kind of initially isolated but as the space grew and as the number of connections grew that
our spaces would eventually meet each other and it's what we've lacked up until recently is a glue
lacked up until recently is a glue that can easily connect these different goals. And that could be
anything ranging from funding to peer review to scientific computing to privacy tech to
any number of different aspects of it. We didn't have like the the the either in the institutional
or the technological glue that would it could would actually enable us to compose each other's
foci and that is slowly beginning to emerge especially in the past few months as these
sites kind of exploded we're now entering a time and a place where a lot of the things that we
wanted to do in the past are now actually realistic because we've also been building for
three years also quietly and working very hard. A lot of people in the space have been working
very hard for a very long time without much recognition. And you know, like at least on our end,
we're building in terms of our marketplace primitives,
these marketplace primitives enable making
an arbitrary type of commitment around,
it could be a peer review
or it could be some kind of scientific experiment,
either in silico or in vivo,
that actually,
that can impact other things.
Like for example, like,
did this experiment happen correctly?
Okay, yes, if it did,
emit this reward from the smart contract,
subject to some kind of peer review from there.
And if that passes, then you can get funding from, you know,
automatically get funding from some other place.
That level of composability doesn't exist yet.
And that's one of the things that we're looking to enable.
Love that.
Serif or causality, if you want to jump in.
Maybe to say some things about reproducibility part,
from my perspective.
I think the whole problem, like, so everything I say
will be from information discovery perspective.
So sorry for that.
I only talk about discovery part mostly.
That's what I know.
So from that perspective, it looks like we have an information organization problem.
So like the whole problem is an organization problem. The research object itself, if we don't organize information with the right semantics, if we
don't structure information in the right way, then restructuring will be a problem.
Then we'll stay with just the narrative itself.
So like reproducibility will take a huge effort.
So that means like the data in the research object
should be shared with the same correct semantics, right?
Like with right attestations,
we should know the whole production pipeline of the data
should be organized accordingly.
Then how the experiments are constructed, Production pipeline of the data should be organized accordingly.
Then how the experiments are constructed, we should know how the environment is constructed.
So it's all about putting more data in the right pipeline.
But it's easy to say so in order to do you need compute in order to reproduce the whole
experiment like that's maybe these guys are doing but like the first step from our perspective is
like discovering the uh environment that the the scientific research is actually did the first time
the scientific research is actually did the first time.
Yeah, I hope it makes sense.
That's what we think about this.
Definitely makes sense.
Really appreciate you sharing that perspective
and kind of enabling that kind of level of access
or trust into that kind of discoverability phase.
DSAI Foundation, welcome.
Would love to have you share what you're building within the DSAI Labs ecosystem.
Yeah, totally. I guess a short overview of what is being built at Laos and Foundation right now is there's kind of these three layers.
Yeah, totally.
There's the base layer, which is our codex protocol, which is how we understand different research objects.
So research objects as composed by not just a publication, but a
publication and code and data and any other artifacts or bits of the research process.
And how we make sure that you can version your research so you can publish your preprint and
then publish your final version, or you can publish even before your preprint, your pre-registration,
you can publish your first iterations of what data your preregistration, you can publish your first iterations
of what data you're looking at.
So we can really see and verify the process of research
instead of just that final output,
which is like the most polished version
and thus also has the most opportunity
for both accidental and intentional fraud.
And then there's this second layer.
So there's that codex protocol,
which is kind of the underlying,
like how we're publishing all of this information
and gathering it into one persistent identifier
or like completely available
and much more reliable way of referring to data
than the current traditional identifier identifier which is like DOI
which requires a lot of maintenance and so trying to use the power of a like web3 ledger technologies
to allow us to have those really high quality identifiers so we're not losing data in the process
in the process. The second bit of this is kind of this higher level like interface to that protocol.
It's like how do we make this really easy for researchers and through that process we're doing
different iterations. One it's just like how do you make it like very familiar like uploading a
pdf or uploading this information that you have as opposed to trying to get on and
use any of the IPFS technologies or any of those. And in that we have our design nodes,
our design publish application. And the additional layer to this is like trying to use the tools we
have now available with AI, available with verification, and all of these different things in order to make
it really easy and automatic to have really high quality data processes. So instead of,
we can have you upload your GitHub, upload all of these places that you're already putting your
information, so it's all gathered in one place. and then the second bit is being able to automatically input all of your metadata so it's not a really arduous process of filling out a
spreadsheet of all 100 of your co-authors but instead you can just upload it and we'll do the
work of finding all of the co-authors just to make sure that we can have high quality data and high
quality research objects um without a lot of the work
of it which is kind of one of those issues where like you don't have compatibility and then the
final bit is we're looking into building a token layer that can help us do verification of this
data through peer review as well as potentially other automated agents or things like that. And we were talking earlier with Causality Network
about that being like having high quality verified data
as a type of review in a way it's not really review,
it's just making sure you can't do anything wrong
in the first place, which does the work of review.
So a lot of the things that we're working on right now
in the ecosystem is just trying to
find a way to make sure that um all of these different ways of handling data um are able to
be communicated across i think we're focused really on that end cycle that publication cycle
where it's like how do you get this available to the world how do you interface with it but then
especially like all the work that causality network is doing around that first set of data like how do you make that data verifiable in the
first place and that co-op hive is doing around making sure that we can do computation over that
verifiable data um and that generalizable computation layer and then the index network
which is like past us even which which is that discoverability layer.
How do you really discover all of the work that people are publishing?
So I think that's in the ecosystem.
That's a little bit of how we see ourselves.
And right now, just trying to really find ways to make sure that we're all communicating well in order to work best together and make sure that we're moving forward also as a functional ecosystem.
So that, again, that like verifiability and the ability to be legitimate to a lot of these pretty traditional institutions makes most sense.
And it's super exciting to be here. Thank you.
Definitely. Thanks so much for that overview.
Thanks so much for that overview.
And I think that's just another puzzle piece that needs to be part of this whole system to then bring more trust and just kind of confidence in all of the different work that everyone is doing in DSi, as well as in more kind of decentralized scientific pursuits to those more traditional players.
Would love for you to also share a little bit
about IOSP in Denver and what this convo overall
is also helping to direct towards or gear up for
that's happening there.
Yeah, so IOSP is a, it's an event that's like the first, it's like a subset of the DCI Summit this year. It's an event that is running, it's like one and a half. It's like, this is kind of the first iteration of it. We had a baby iteration last year
that many people were at called SciOS,
which was a really workshop heavy orientation.
And this one we're trying to learn and iterate
and start with day one,
which is a lot of researcher presentations,
trying to get into the use cases
and needs of these researchers.
And then the day two is really
heavily focused on workshops and then day three is really work focusing on all of us um trying
to figure out how to work best together and that can be both on the level of like organizations
and on the level of technical interoperability i think the the goal is to see where we can start plugging
the holes in each of our different functionalities
so that we don't just have, for example,
causality network who has verifiable data,
but then we can actually open and use that data
for a researcher potentially.
So making sure we can like plug each other's gaps.
IOSP is hoping to happen every year so the goal is that
this isn't just a one-off like show up and see if we can hack together a little bit of an intro op
but really holding the space for the work of connecting between different platforms to be
actually possible so that's the that's the main goal that's going on at IOSP. So everybody's invited from the February 22nd
to February 25th.
And I can send a link to that afterwards,
but mainly just like using that space
as a place to collaborate is the main goal.
Definitely need more of that.
So looking forward to the future iterations as well.
I guess gearing up for that,
maybe if we each go through,
okay, we've done an overview of this is what each of our projects are,
but possible ideas we might have of,
hey, this is what could funnel into us or this is what we could funnel into.
Possibly also any other just kind of requests across the board.
Maybe it's some type of feature opportunity that you're seeing,
but just isn't the core focus for your project or organization right now.
But you could see someone else in the ecosystem possibly entirely focused on that to
fill in that gap. That might be helpful just kind of to set some of the stage of what are some of
those pipelines in, pipelines out, or just other opportunities to bridge more connections.
Ellie, feel free to keep going as I think you may have spent possibly a little bit more
time thinking on this than all the rest of us.
If you have some thoughts, top of mind, and then we can go through the rest of the projects
as well.
Yeah, I think one goal that I have maybe with this call is also to evaluate feasibility.
Like I know Labs has been working on getting an API up and working, which think one goal that I have maybe with this call is also to evaluate feasibility.
Like I know Labs has been working on getting an API up and working, which is one generalizable
way to do interoperability.
So I think I'd kind of like to hear from each person, like partially one, like what is the
most powerful or important collaboration that you want in that gap of like, hey, like it
would be actually like
this one, if I just interrupt with it, with this one person, that would be really helpful for the
like services I can offer. And the second bit is like, where are we all at in terms of that,
like the work of interop? Is it like, would it be, oh, like, here's like, help me explain my API,
or is this going to be something where we need to work together
to figure out like what it is to interop
and what steps we each need to take
in order to make sure we're in that position.
Yeah, like the feasibility aspect
is definitely a huge piece of it too.
From my perspective, and this is mostly with talking with some like monolithic collaboration
usually doesn't, it usually hasn't worked
because it's just too much effort.
Like there's a number of blockers there.
I mean, one, it requires a ton of effort from both parties.
Two, it, it requires a ton of effort from both parties to it.
Um, you like the two, the two projects or protocols or whatever are not always built to work with each other, like they're both in quite different ways.
Um, and then, you know, there's like all this other stuff that people have to do.
And so, you know, it usually oftentimes doesn't end up happening.
oftentimes doesn't end up happening.
And so I think that what we need are like smaller, more modular places where we can
And so.
be like, oh, here's an opportunity where I can use, you know, these at last tech stack
or where I can pull data from that causality network created or where I can do some agent
discovery aspect of Seraph of index networks protocol.
We need like smaller, more modular building blocks that can also be used by other people.
And I think that that would help increase collaboration across the board because whatever
is created between two groups for collaboration, if it's small, it's modular,
can also be used by other people who can improve on it
and inform its improvement in the future.
And yeah, I think that's been...
Well, we'll see if that's what ends up happening,
but I think the reason for lack of collaboration up until now,
not lack of collaboration,
it wasn't lack of desire by any stretch of imagination.
It was one, a lot of our projects were premature and they're maturing rapidly now, but also
two, the types of things that we were talking about doing together were very extremely time
consuming and would require serious work on the part of both parties on the order of months.
And I think that if we moved away from that model and towards like more yeah like modular smaller more modular shareable pieces
i think that that would increase cooperation dramatically 100 i totally agree with you maybe
yeah exactly and that was i was thinking while listening. So smaller, modular, reusable pieces, like are these software, are these data,
or maybe we can think like how it didn't work before, right?
Like how we couldn't interrupt previously and how it's possible today.
So there are, let's say, two pieces.
The first one is the data, the second is compute.
And with using these tools, we want to collaborate and communicate.
and interoperability is a basic terminology for this.
And interoperability is a basic terminology for this.
Alright, so first as the industry we try to be interoperable over data.
We try to set data standards like the schemas and all of it.
Of course it could work but mostly if you contextualize the data based on one single standard, you end up with some
kind of monolithic thing because you organize the data for very static thing.
So interoperability based onualizing it was kind of hard
at that time.
For science it was actually possible, but for some reason it didn't accelerate.
So how about, like as Levi mentioned, like smaller modular components and what are these
components?
Are these again the data i
think these are kind of functions and if we reprint these functions maybe we could rephrase them as
agents and since autonomous agents they are already composable right agents can talk to each other, they can communicate, they can come up with new methods for open-ended tasks.
So they're already interpretable, but in the functional layer, not at the data layer, in the functional layer.
I think that's a huge opportunity for interpretability and requires us to redesign our assumptions for interoperability.
Maybe like we should, as the most experienced people in the space, define like what are
our responsibilities for this agent ecosystem.
For example, as index network, we are building the discovery part.
So we will be building the indexing agent for example maybe the other
part will be levy building the compute scheduling agent right so i think in six months in one year
these agents and these roles will be clear and by just deploying our agents and making the data just available to access and accessible in any form, like not just build on one standard. Of course, standards are good, but maybe like we can embrace more than one, more than five standards or different ways. Like it will be more flexible, what I want to say on that one.
more flexible what I want to say on that one. I'm very optimistic with this agent paradigm
for interpretability in general as a summary.
I'm wondering Levi, would you mind like defining a very specific module as well that you'd be
interested in?
Because I think the way I'm thinking about it, for example, like a module of collaboration that I like that we,
Causality and I were talking about or and Labs was talking about in the hour before this was like.
If Causality has data that they have verifying and we want to find a way to potentially send that data into a
community that then has absolutely verified data as like one baby module and it's like one tiny
touch point so i like i super agree with that like we need actual tangible graspable collaborations
as opposed to like making a collaborative vocabulary for scientific data because that that's crazy and like a big problem
that a lot of people are focusing on not me um but i'm wondering if you have any specific ideas
of like what is a what is a biteable chunk of collaboration that you'd be interested in
dealing with on the 25th yeah i mean like just one thing that's come up with like actual multiple different parties
is like, for example, exchanging like an attestation, which is, you know, there's a lot of protocols
that use this Ethereum attestation service.
It's just like really just a way of signing and testing to data on chain.
It's like, for example, exchanging an attestation for an NFT or ERC 2721 or an exchange of
an attestation for ERC 20.
It's because you can say, for example, like,, that, uh, the output of the complicated, the, uh, you know,
when I put the, you know, the causality network for this example, when I take this, um, you
know, uh, when I put this, you know, a mixture of reagents or whatever, you know, uh, like
a centrifuge or something, I think I'm just making things up.
Um, not reagents, but you put this like liquid in, in a centrifuge and something. I think I'm just making things up. Not just, but you put this like liquid in a centrifuge
and, you know, this is like the result afterwards,
you know, this is being attested to by a TEE.
Like, it's like, oh, okay, you know, you attest to that.
You have that like attestation on chain.
Now we can exchange that for an NFT
that like gets you membership into the, you know,
correct centrifuge club or something like that.
Or you can say, oh, okay, you ran that experiment,
you zone laboratory resources, that costs some money.
Here's this ERC20 token to help offset the cost
of having run that experiment.
It's like, that's the's kind of like the kind of
six smaller building blocks that I've been thinking about also
like a peer peer reviews, another one, right? You can attest
to something being like, oh, okay, I said that, you know, I
test that this expert, you know, that this paper is correct, or
this theorem is correct, or whatever. It's like, oh, okay,
well, you know, we want to like, automatically reward you for
your work of making that attestation. Here's this ERC 20.
And so the primitive that we that we have to like automatically reward you for your work of making that attestation, here's this ERC 20. And so the primitive that we have like in our code base
is like it would be something like ERC 20 for attestation
or ERC 721 for attestation.
But we can also do ERC 24721 or, you know,
we have like almost an arbitrary mixture of things
because we've built things up
from like very, very small building blocks.
I believe that was the type of thing I was thinking of.
Barrett, do you have any additional thoughts along these lines?
hopefully i don't get rugged again um no no i 100 agree with what ellie's saying about
Hopefully I don't get rugged again.
doing small things so we've actually done something together and like i haven't got much
to add in terms of what was actually discussed but i think it's super cool we're all working on
different aspects of even just the reproducibility crisis and we're tackling it from different angles.
Yeah, super excited for the ISP event.
Yeah, nothing else specific to mention.
Cool.
Ellie, leading up to this other event, are there any other kind of pieces in addition to that feasibility aspect that you kind of brought up and then some
of these more like let's do things in more modular types of ways to get some of these different
integrations started and like at least explore some of these different pathways. Are there any
other kind of things that have been top of your mind just prepping for that event and thinking in this cluster?
I think the main thing that has been coming up around just like organizing more workshop type
events is that like you can spend the whole workshop figuring out what you're going to do
but then you leave and if you leave without doing the, then a lot of times it is gappy.
It doesn't fully happen.
I think this is part of the reason why we want IOSP
to happen multiple times,
because likely the first event will be kind of figuring out
what we want to do.
But I think one of the things in prepping for that
and doing as much as we can ahead of time
is just each figuring out, yeah,
just the specifics of what would be most valuable
um so then once we're in that space we can start with a really informed place of like these are
the things that we're all looking for and we can kind of coalesce on little things that we can do
um so i think i'd love to hear from like everyone everyone here of the specific things that might be valuable.
And then anyone who's listening as well,
just dropping in kind of ideas that you guys have
around what gaps there are.
I think mainly I've been thinking about this
in terms of the research data life cycle,
which is just like there are times where you,
like you're gathering the data, you're verifying the data,
you're evaluating the data, analyzing the data,
publishing the data.
And so finding ways to create a little pipeline across
is like the ultimate vision.
And figuring out what we're going to do in that day
is in order to be like,
use the time that we have best is my main goal. So just that we go in there informed on what we need and what we can do.
So I think that's mainly the things that are most interesting for me.
synchrony. Definitely. I know before you were bringing up kind of the positioning of an API to
Definitely.
be one way to kind of connect different things together. Maybe it might be useful for everyone
to go through if there are any standards they might be following or any kind of public either documentation or if an element of their product is live, like a direct
way to kind of start playing with some of those different funnels or just kind of thoughts on what
some of those next steps of communicating from each of our sides of like how we might be able to integrate and just from a technical perspective.
Yeah, Ellie, if you want to kick it off and then we can kind of go through the list.
Yeah, sure. I think the understanding that I have from a technical perspective
and I think Alkban you're here, if I'm like messing it all up, please request to speak
and you can add in information.
But my understanding of what we have
is an API to the Codex protocol.
So ideally we have this option to publish
to the Codex protocol, and then you'd see this visibly
on the node, as well as the opportunity to read what we have
on the protocol and navigate that data
in any way you would like to.
That's the main work I think we've done on interoperability.
The only other like broader level information
is like everything's open source.
We're still working on making that functionally open source.
So like it is open,
but if you can do stuff with the codex protocol that's something to
work on as well the only other thing i might mention here is that we are working with ceramic
and using the um i'm forgetting the name but there's, we're using ceramic as our kind of base level understanding of the data structure that we're using.
So this gives us the ability to use streams.
So it might be easier to collaborate and integrate
with other people using that situation.
So depending on what everybody's agreed on
agreed on in terms of standards, that might be something to note.
in terms of standards, that might be something to note.
Great.
Merrick, welcome.
So good to see you.
Yeah, great to see you.
Welcome.
Yeah, it's good to be back.
I took a holiday break.
Glad to come back and see D-Sign Mike community, like vibrant and having amazing discussions.
I apologize for missing like the first half, but I actually did have, you know, aside from
good wishes and things, I actually did have a comment, which is why I requested to come up.
You know, I, you know, like a lifetime ago, you know, used to be like an academic researcher,
did like a few postdocs and things like that and we were dealing with like a lot of like um kind of like health related data if that makes sense right
health sciences and you know medical records are you know it's sort of a strange beast to begin
with right but like you know kind of case in point um you know um you know like health care
provider a would describe like a patient as an inpatient, right?
Meaning someone who's like staying at the facility, example, hospital, and then like
healthcare facility B would define a patient as an inpatient or outpatient, right? Which means
somebody who's just like, you know, comes in and out for whatever treatment, right? And, you know,
I think that really kind of, they couldn't even agree on like the most basic, you know, terminology in healthcare, right?
So apologies if this was already discussed earlier in the space.
But I mean, I think that there's two kind of levels of, you know, interoperability, right?
One is like on a technical level, right?
Meaning like, can I connect your API?
like, can I connect to your API? Are the data structures, you know, the same? Are we using the
Are the data structures, you know, the same?
same, like, Indians in, you know, in describing numbers, these sort of ideas? And then there's
also, like, a semantic layer as well, right? Like, is, you know, when you're talking about
some phenomenon or, you know, scientific term, like, are you using that in the same way that I
am, right? And, you know And we never really kind of completed this research
because it's a lot of work,
but I would just like to point out
that there's possibly even more important
than on the technological level of interoperability,
there's also like this semantic interoperability, right?
Which I really think is interesting,
this idea of letting like LLM agents
do some of this like knowledge translation communication.
But I think, you know, having, you know,
grounding term, like terminology truth, right?
Like semantic truth in some agreed upon,
you know, I guess the tool is like an ontology, right?
And, you know, just a thought for, I guess,
the panel if anyone's thinking about that.
I'll pop in for a second.
We have a workshop focused on this largely
in the second day of IOSP called like data visitation.
And the main thing that people are kind of landing on is like, hey,
we need to know how to communicate across projects.
The semantic interoperability problem is like the problem.
So yeah, I agree.
Just on that point about the semantic things and the ontology, I don't know if anyone here
is familiar with IEML, information economy metalanguage?
I myself am barely familiar with it, but I came across it quite recently and I think
it's super interesting and applicable to a wide range of science related things.
If anyone's interested in that, I think definitely do a deep dive into it.
I'm not fortunately technical enough to understand it.
I think this is such a good point.
Even the work that we had previously done with TalentDeO of even just defining what does it mean to be part of a
DSI organization. Every single organization had completely different terminology of what it meant
to be core team. And that's something that I feel like should be even more basic that we could agree
upon, let alone at some of these deeper levels or just kind of a wider range of different data types or
information flowing into different systems. I know we've been talking a lot about just
like the flow of creating standards but also needing that behavior to align with those
different standards. Otherwise just creating the standard without that behavior or
alignment or adoption, like that doesn't really work. And also, if the standard doesn't
align with that behavior, then that doesn't work either. And people are just operating in a
different type of way. Yeah, Merrick. Yeah, sorry, I'll jump in with an old man story again, right?
You know, these sort of ideas of like creating standards, I mean, you know, I guess there's no other way, right?
But like, you know, there's an old programmer story that goes kind of like this, right?
It's like, you know, a programmer sees, oh, there's, you know, 100 different ways of doing this thing, right?
So I'm going to create like the one ultimate way of doing it, right?
And this programmer goes off into their programming cave.
And before long, now, before long, there's 101 ways to do this thing, right?
So like, you know, standards are kind of like paradoxical, right?
I think as you pointed out, Aaron, like, you know, people have to like coalesce around
this standard and, you know, creating new standards isn't necessarily the way, which is, you know,
I appreciate, you know, the, the earlier comment that, you know,
there's already, you know, different frameworks and, and, and, and,
you know, things being created. I think it was called I something,
anyway, it was mentioned, but yeah, it was just a funny story,
a bit of like a paradox about standards, right.
As you try to create the one standard and you're just kind of adding to the mess.
So not to say that that doesn't need to be done and it's not important.
It's just something maybe to keep in the back of our minds.
Definitely. I think that ties into what Sarah, if you and Lovey were saying before,
of like, let's start with these modular things, get different things up and running, rather than trying to, like, take on this entire beast.
Yeah, Serif, I think you were going to say something, so keep going.
Yeah, that was great to hear semantic interpretability, word, because, yeah, that's great hearing this again.
And it was kind of our utopia, right?
For, let's say, for 20 years, 30 years,
semantic apps utopia are like a end goal with semantic interpretability.
Finally, to be able to understand each other,
to communicate in an accountable way,
like concepts, abstract things or scientific
findings to communicate these so like semantic like and so yeah thank you for liking my comment
and i want to add on that one a bit so why semantic interpretability didn't possible
because again like we couldn't reconstruct the things in context again.
So we have to design these reconstructions.
We had these interoperable data schemas,
but we had to redesign our own apps or tools according to these schemas.
But now finally we have semantic processors we
have AI AI can process semantics that's a perfect tool I think like we have to
recheck our assumptions and historical assumptions like do we need standards
for doing this or what kind of standards for the data do we need for example
like validity okay the
validities are like still important it's going to be important we need to know how it's produced
but maybe we shouldn't be caring about the schema anymore because we have you know like these
assumptions are going to change uh but like yeah Utopia, I think, is very close with this semantic processor's so-called agents.
Let's do it. Keep working with Utopia. Merrick, yeah.
If I may, again, you know, I think, you know, just to repeat, and I agree with you there, again, you know, I, I, I, I think, you know, just to repeat and I, I, I agree with you that like, you know, agents, LLM based agents, which are not the same thing, right?
I think that, you know, our industry has equated, you know, LLMs wrapped in, you know, some code and APIs is like an agent and it's not like, it's not the end all and be all of agents, right?
And it's not like, it's not the end all and be all of agents, right?
Secondly, you know, if I've created, if I'm using an LLM, right, that's based on, you
know, some different, you know, training data than yours, right?
There could, you know, semantic misunderstandings could potentially creep in there. So I think in addition to that, the LLM technology is subject to things like hallucinations and
so forth.
So I agree with you.
I think that a lot can be done with that.
It's just maybe like a word of caution that these sort of things could creep in or occur.
As we're kind of continuing down these paths, I've always seen DCI's being kind of the second
iteration of a lot of the open science movement.
Are there any organizations, initiatives, maybe attempts at kind of creating these different layers of coordination or standards that each of you might be have followed that that we should just kind of keep in mind as examples to learn from?
Ed welcoming you up here in a moment as well.
Ted, welcome.
Do you have some thoughts on this?
Tasking, there's the RFK confirmation hearing,
which is pretty important in the United States right now.
So I've been trying to monitor both because I'm going to be speaking about that later.
But I mean, I guess my general thoughts on the general topic is, and I often mention it is,
you know, how do we really connect DSI in the overall desai because people don't know i come from the regenerative ag
movement and we're not just about ag we're about understanding health or understanding what is the
relationship between the soil biome the nutrition and our food and our gut biome and our overall health. What is really health? And it's a movement that's
been going on for a long time, does a lot of science, has a lot of data, but it's not really
Web3. It's not really token. It's not really, you know, it has the same problems that all decentralized sciences had.
And it works on some ways of solutions of that.
And I think there's a lot of potential to connect the different forms of DESAI,
whether it be DAOs or whether it be tokens or Web 3 or Web 2 or Web 1 or just networking designs and media
designs of communication and working out these situations so that we understand in regenerative
ag especially because it was kind of the roots of this work. It was how do you network?
And this was even before computers.
And how do you really understand what everybody else is doing
so that we're not repeating, so that we can build on knowledge,
so we can do proper peer-to-peer review,
so that we make sure that it benefits the good of all
and isn't necessarily trying to just make money,
which I think we've all realized in the science movement has really changed.
You know, it's a lot of corporate interest in making money instead of what it actually does for the good for the people or the planet.
what it actually does for the good for the people or the planet.
And so I continue.
It's not easy to how we're going to network properly because, you know,
as much as we call ourselves decentralized,
the most important thing is that we are very well connected
and very well understanding of what we're all doing and set some standards for it. You know, if we
find an area of science that, you know, maybe I believe in researching science, but the
implementation then has been a lot of problems. We've seen that in the drug world. We've seen
that in agriculture, in the pesticide world.
And that's why, you know, in regenerative agriculture, we are not strict because we realize the system has to go through a transformation.
if when at all possible to get things out of our system like a lot of synthetic chemicals and
even some biological technologies that we can't be sure about their safe and effectiveness
and so that if the more we can just understand the workings of nature which we realize now actually has a lot of the answers without having to necessarily go
into the synthetic world of these things. So I'll land it there. But as I said, I'm sorry,
I haven't been able to listen closely, but I wanted to give a little quick feedback
before you closed. And I will re-listen to this space later on.
Yeah, thank you for joining into the convo.
I know Sarah just had to hop off to another meeting,
but definitely check out what they're building
over at Index Network,
really focused on that discoverability layer.
Nowhere at the top of
the hour, so if other people have to hop off to other meetings, thanks so much for joining.
And for any of the speakers still up on stage, would love to kind of do one last go-around of,
if you have a couple minutes, of kind of what you might see as some of those next steps,
couple minutes of kind of what you might see as some of those next steps, maybe a modular,
smaller type of integration idea that we might be able to work on over the next few weeks
leading up to that IOSP day during the ETH Denver timeframe.
I do have to hop off, but it's been really, really great talking with everyone here and
just hearing the updates.
Yeah, I'd say like just quickly work before I leave, regarding the workshop, I think definitely
like having like small achievable goals as the outcome of the workshop
would be really, really beneficial.
Just so that we can like use that,
I can make the momentum of accomplishing that,
those smaller ones to then accomplish bigger things.
And I think that there's definitely a lot of,
there's so many things we can do together.
We just have to kind of nail down a few use cases
and just kind of like really push together
to make them happen.
Absolutely.
Are there any immediate use cases that jump to mind
that might be top of the list
for people to think on in the meantime
or more of a call to action
for us all to be thinking on that?
I think a lot of people here are concerned
about the reproducibility crisis.
And so anything involving like rewards for verifiability, I think would be really, really
useful. Definitely. Eric, causality, I feel like that might be a good lead-in to what we're
working on over there.
Yeah, absolutely. And for the small, tangible thing we want done during IRCAP, it's just
we have one device that's just doing one basic experiment where we've
got an API so someone can kind of externally verify that it came from our network.
So that's our kind of goal.
Awesome.
Desai Labs, Desai Foundation, Ellie? foundation ellie yeah i think our one goal um for now is just to have like one person use the api
and tell us how it's broken um and also and just to be able to see how we can best um shout the
word out there for any of the data that's getting generated. Love it. Amazing. Merrick,
any closing thoughts on this conversation or maybe leading from some of your other experiences
trying to build out things across data and standards and all of this as well. Yeah, no, I mean, from what I caught of it, you know, like others,
I'll circle back and listen to the half I missed.
But, you know, great to see everyone here again.
You know, fantastic discussion.
Aaron, love what you mentioned about verifiability,
reproducibility of, you know, results and things like that.
And I think, you know,
a lot of that comes, I've got a lot of thoughts around this, I'll make them super brief, super brief. You know, it's, you know, the crisis is there because of misaligned incentives in
traditional science, I think with DSI, DSI can actually fix this, right? I think it's just a
matter of like the will to fix it. And to say something like, you know,
if there's some D-Sci DAO allocating X dollars
to some, you know, experiment, you know,
allocate Y dollars to like reproducing that afterwards, right?
If there's a positive result.
So like, I mean, I think we can do it, right?
So great topic could be a good one for another day.
But yeah, that's it for me.
Amazing.
Yeah, we will be continuing the conversation
in person during IOSP around ETH Denver timeframe. If you go to IOSP.io, I believe that's what it is.
Ellie, correct me if I'm wrong. IOSP.io. Cool. Great. Yeah, Go sign up there. Join to be part of the conversation and just working
towards making all of this reality. We'll be back here again next week, same time, 5 p.m. UTC,
continuing with more conversations. We might have some folks on either next week or the following week who are building out an entire blockchain for DSi. So I think that just lends into the whole conversation about how can we
bridge some of these connections, really kind of tap into the leading opportunities of these
technologies to really advance DSi and science forward. Awesome. Well, thank you to everyone for joining in.
Let's keep thinking on all of this. If you have other ideas for these modular components or
different ideas that you would like to see come out, feel free to message here at the dsignmic
account or directly to my profile, Erin, or if you resonated with something someone
set up here, reach out to them, give them all a follow.
All of those connections are what makes all of this possible moving forward to just make
sure it's having the impact we desire as well.
Awesome.
Keep building cool design stuff and we'll see you back here next time.
Thanks so much, everyone.
Thanks so much. Take much, everyone. Thanks so much.
Take care, everyone.
Bye.
