Projects
![]()
No projects found
Investors
New
DAO Treasuries
DATs
Lend / Borrow APRs
Chains
Categories
Spaces
AI on NEAR Protocol ft. @nansen_ai
Recorded: March 27, 2025 Duration: 0:59:18Space Recording
Short Summary
The discussion highlights several key developments in the crypto space, including strategic partnerships and innovations in AI and blockchain technology. NIR's focus on dynamic sharding and decentralized confidential machine learning showcases advancements in scalability and sustainable AI models. Emerging trends such as AI-powered tools for smart contracts and agent-to-agent payments are also explored, indicating a shift towards more automated and efficient systems. Additionally, initiatives like the ZetaGash Ed's hackathon and NIR's ecosystem growth demonstrate ongoing efforts to integrate AI into blockchain applications.
Full Transcription
Hello, welcome everybody.
So we are just kicking off these Twitter spaces.
Welcome, camera.
We also are seeing our guest, Nansen, that is also with us.
Our co-founders are also in the audience.
So give us a couple of minutes so more people can join,
and we will start talking about AI on NIR.
So you help us a lot if you share these Twitter spaces
with more people so we can reach a higher audience.
also being able to have
to have spread the word about what is happening on AI inside of NIR.
So let's wait a couple of minutes and we will start talking about this topic.
Meanwhile, I think that we have here both of our guests for these Twitter spaces.
So welcome, welcome.
First of all, we have a Cameron, Cameron that has been contributing to the NIR protocols since, well,
a couple of years that a year inside of crepto is like
more time than it seems so welcome Cameron how are you doing
I'm doing great it's good to be here thanks for having me
awesome so also we have with us Nansen can you hear us can you are you able to
speak
Hi, it's already speaking for Nansen today.
I had the research team at Nansen.
Yes, I can hear you.
Okay, so welcome, team.
Well, we can start with a round of introduction,
so we can be talking about what is happening on AI.
I think it is really trendy at this moment to talk about AI,
and everybody wants to talk about AI on the ecosystem.
So, yeah.
But there is always like this need to show the value that is being presented to the community.
Because AI can be many things, but at the end, what we're looking for is to deliver something to the community.
that can be used it, that can improve their lifestyles
that can improve their business processes that they have
inside of their organizations.
And the intersection that we are having right now
among NIR, an AI,
It is inspired by the long history that the near co-founders have inside of the development of AI.
If we remember, both of the co-founders have been working, early analysis has been working on AI since.
almost a decade. And right now they are focused on creating this intersection among
public transparent network on blockchain that is being supported by AI. So, well,
more than that we can start with the round of introduction. So Cameron, do you want to introduce yourself?
Yeah, sure.
Hey, everybody.
My name is Cameron Dennis.
I've been building in the near ecosystem for about four and a half years or so.
Been in crypto for about eight, but really been focused on AI for the last year and a half.
And I'm on the near AI team where I lead a lot of partnerships and integrations.
primarily, currently focused on infrastructure up to now.
This infrastructure includes building the private MLSDK with Follow Network,
so you can run any Docker image inside of a trusted execution environment to verify privacy
and also making sure that the agents or model that you're using is the one that you think
you're using.
So working on that for quite a while, I've been getting a lot of actually the person to get
get all the compute to provide.
So broken a lot of compute deals.
And now I'm currently working on some agent that can audit other agents.
We call that an agent auditor.
And that is really important for security.
browser agent, so an agent that can kind of navigate your browser and click for you.
There are a bunch of other teams in the ecosystem.
They're building really cool things that I work with every day, like vectors,
your private databases, so your agent can actually ensure that it knows about you to do helpful things.
In short, I'm pretty much like the only non-core engineer at the NIR AI team,
responsible for integrations and partnerships.
Yeah.
Fantastic. Thank you, Kamara, for spending an hour of your time for talking with us.
I think this will be an awesome opportunity for sharing what is happening on AI.
Nancy, it is you, Enri, do you want to introduce yourself?
Yes, my name is O'Reilly Barter.
I work in the research team of Nansen.
And what we do, we use on-chain and off-chain data to advise investors, traders, builders,
protocols on different topics.
And we've also, of course, we're also including AI into this process of data insights and manipulation.
Okay, perfect. So, well, and personally, I'm Alan, has been working on MetaPol since three years and a half at the moment.
And right now we are supporting different initiatives to help to decentralize the networks.
Obviously, AI is something that we are working on during this year.
We hope that we can share more information about what is happening.
So just to warm up this topic, so...
I want to ask for that you can comment.
What do you think, how do you think that AI is right now
shaping the future that we are working on blockchain and Web3?
So what do you think are the main chains that are being proposed by AI,
by being able to?
to manage the blockchain interaction that we have with the users.
Cameron, do you want to talk more about this?
Yeah, for sure.
So I'm going to kind of bring it from the first user perspective,
and then also the developer angle for the users.
We haven't seen this take root 100% yet,
but there seems to be a lot of promise around AI making it easier
for people to navigate Web3.
And because as everyone here knows, that's been in Web 3 for Quoteau or Crypto, it's pretty complex.
Key management is complex.
Naviting all these different user interfaces gets complex.
You borderline need of background in finance in order to make money in Defi.
Obviously, there's some really great teams that are streamlining this, making it easier.
But a lot of people still confused around the concepts of staking and liquidity pools and everything else.
So I do think AI is going to streamline the user experience for people,
where they can communicate with their assistants in natural language,
and this assistant is able to perform actions on behalf of the user.
Now, I will say there's a bunch of issues with this,
primarily around security.
And so I do think that is one thing that could be improved by AI in crypto.
So, yeah, people, everyone in the world has money wants to make more money.
And I think AI will, you know, do a pretty good job at helping people make money.
But the problem with that is around privacy, security, and privacy.
Gosh, it's really big because wallets need to adopt this as well.
You need the user interfaces to actually get access to those users.
I think like every wallet in the world will eventually have an assistant that is able to navigate metapool easier than users do.
So that's like the B to C.
And then there's the B2B play, business to business.
where if agents actually do end up automating the backends of businesses,
agents will have to communicate and pay other agents.
And these agents are likely going to want to pay in stable coins
because it's a peer-to-peer transaction or age-to-agent transaction
that costs fraction of a penny rather than having to settle on stripe
for 2.8% of the transaction.
And so I think the B2B play, business to business,
is much more focused on
driving costs down for users and businesses, which is the best way to, you know, provide value, I'd say, a one of them.
So, yeah, B2B and B2C.
You mentioned something that is relevant because I think that we, for us that has been here on the ecosystem for a couple of years, there is always this need.
for having a better user experience that the one that we currently have.
Because being honest, it is hard for a new user to get inside of what is happening on blockchain.
I mean, passing through a wallet, funding this wallet, making something inside of your...
of your wallet results on a very long journey where you can lost a lot of people in the middle or getting
in worst cases, and there is like loss in funds or something similar.
So I'm totally agree with what you mentioned, that AI compute and can make this difference for the users.
And we hope that this also can be, can help us an onboarding tool for having more people inside of the Web3 ecosystem.
So coming back to the Enhan team, so.
The same question, what do you think about this?
How do you think that AI will improve the user experience that we have right now,
who will improve the ecosystem that we are working for
when we have this integration with agents and AI.
Yes, I guess it's a little bit of a similar framework as the B2B and B2C, maybe it'll
try it a little bit differently.
You could have, for example, AI as a revenue on the blockchain.
So if you look at existing protocols, there are protocols that are like marketplaces that
offer other data that you can optimize or build on or just really AI endpoints, API endpoints.
So thinking a bit of as a marketplace where you can sell different services,
could be agent services as mentioned before.
So that's a bit like the revenue part and crypto or blockchain would be,
I would say like the infrastructure to market that.
The one that is also promising is probably correspond to your B2B framework
is it drives costs and down and improve productivity.
So in the end, protocols or chains or just underlying kind of businesses that are trying to achieve something.
And for example, at Nansen, we use it to be just more efficient, effective in different processes to do our job.
So that's the second one. And probably the third one,
is around infrastructure because of the whole tokenomics framework.
You can sort of manage, for example, resources like CPUs, energy and so on.
And we've seen certain protocols try and using tokenomics to do that
and compete, for example, with cloud computing.
So yeah, maybe a free side framework.
very interesting, very interesting. I think that there is, there is still work to do with to have all of
what we're thinking on this kind of development. So let's see how this evolved. There is always
like we have like this innovation course that we have something amazing that comes up in the middle
of long journeys of work done 30 months or years that we can have. So surely,
this year will have surprises. Let's see we can have like the golden apple that we are looking for
inside of AI during these years, but surely there will be changes on how are we doing the things
inside of Web3. So Cameron, I want to focus like in a specific question for you. We know that every
every blockchain is taking somehow the AI narrative.
We mentioned that near by its founders,
they just started working on AI since several years ago.
Can you tell us how does NEAR differentiate itself from another blockchain network that are interacting AI?
What makes NEAR like the ADL ecosystem for AI deployment?
Yeah, absolutely. So as you kind of highlighted, and you did start as an AI company. And so this has always been kind of a long-term goal. So the first goal was to build a protocol that can scale forever because assuming that agents are operating on top of some sort of protocol for transactions, it needs to scale. It cannot go down. It needs to actually be able to scale forever. And we do that through dynamic sharding. So, yeah.
As more demand comes to the network, whether through humans or agents,
that more block space can be created through the concept of sharding,
and more validators end up coming to the network because you need more block space to validate those transactions.
And we've built a really easy and flexible way to manage this.
So the very ground level needs to scale.
That's one.
Two, these agents that need to interact with any other protocols, they need to be chain-abstracted.
And so in order for chain abstraction to operate properly,
one, you do need a scalable protocol,
but two, you need a programmable multi-party computation network
that we call chain signatures that allows an agent or an AI
to interact across any blockchain that supports ECDSA signatures,
which is an elliptic curve designed primarily, well, it's not designed for,
but that is used by pretty much every protocol that's not Solana and Trond.
And so the reason why this is important is because users and especially agents are not going to care about what database they're working with.
They need to be able to interoperate across everything.
So second, you need to be able to manage these multi-chain transactions easily.
And we enable interoperability.
not at the virtual machine level,
like an Ethereum virtual machine or Solano virtual machine,
but at the signature scheme level.
So that enables much better interoperability.
So your agent can interact with Bitcoin or Dogecoin or XRP or anything,
because at the end of the day,
these are just distributed databases.
Users do not care what database,
their favorite apps are deployed to.
So with near you get that inherent multi-chain interoperability.
And then three, on the agent piece,
Near AI is building agent hosting network,
payment layer and an assistant.
We're working with a lot of ecosystem teams to do all of this,
by the way. We are not doing this alone.
To really sort of supercharge Asian developers with free inference,
which is a really big deal if you're an actual entrepreneur
building agents, less of a big deal if you're just some hacky,
you know, developer that is just building bunch of side projects.
But free inference is a prerequisite for, or like a huge value add to agent developers that are like founders.
Two, confidentiality in your prompts.
So if you're actually building an agent that needs to get access to personal information to actually become useful, you need confidentiality.
So we're enabling this through your hosting network.
The date for actual confidential inference is slotted for later, sorry, like mid-April.
So that's coming out then.
And then three, discovery of your agents, though you can actually gain demand.
Yeah.
So with the near ecosystem, we have about 25 million weekly active accounts.
A lot of these projects that manage all these users,
they're all going to sort of adopt a near AI assistant that is white labeled
that calls the agents that are hosted in the hub.
And why is this important?
Because it drives demand for agent builders.
Agent builders want two things.
They want low operational costs and they want demand.
We're providing them with both.
So it's pretty straightforward.
The ecosystem that actually cares about AI,
that's investing in AI.
We do all these Twitter spaces and hackathons
and all these ways to get involved.
But in short, you actually get an ecosystem
that has been prioritizing this for a while
and has all the infrastructure needed to scale.
Yes, I think that probably I will following this idea that you mentioned Cameron.
We know that in February it was a $1 trillion agent hackathon,
and also right now is happening the ZetaGash Ed's nearly intense hackathon that is also using AI agents.
How hard these hackathons going.
Well, if you can mention something like making an open invitation to the people,
and I don't know if you're directly related,
but probably we can give a wink to what is happening on this side.
Yeah, hackathons are great, in my opinion,
to get developer feedback and awareness and build culture,
but they're really not the best way to build useful agents.
It's a good way to build the culture around people building useful agents.
But I'm not incredibly optimistic about useful agents coming from hackathons.
Most useful agents are coming from businesses and founders that are very clear about the value that their agents provide.
And that's something that I think is seriously missing in the Web3 space.
Lab 3 has a tendency of being very hacky.
What I mean by that is like independent developers building interesting stuff,
but it's not really serious.
It's a bit more focused on just like gimmicks and culture.
So what we're running are like useful agent hackathons.
And so for the useful agent hackathons,
you need to prove that your agent is useful
by providing some sort of benchmark or evaluation framework
saying, hey, my...
social media management agent is for posts are performing
better or worse than a human doing these things.
So we are really focused on supporting projects
through hackathons and things,
but primarily through demand and inference
for building useful agents.
And hackathons are a good way to provide that Taha funnel,
but at the end of the day,
it really does take dedicated support long term.
So I actually have a call later today with Claudio
to talk about how we can further engage these useful agent developers.
So right now, we just have wrapped up a useful agent's hackathon in San Francisco.
There will be another one that runs in New York.
And I believe one online, but to be totally honest, I'm not as focused on hackathons these days.
I'll help where I can.
But the best place to kind of learn more is either near Devhub or near Foundation's founder's success team.
And they'll be coming out with the announcements as they're ready.
Yeah, so totally agree on that.
I think that, yeah, it can be a discovery tool,
but not the long-term process that we can have for developing agents,
where we can involve funding, funders, and other tools.
But, yeah, totally agree on that.
And also, well, on the Nansen side, we...
Well, we have been mentioning that it is challenging to build AI power tools.
It is like even any touch development have its own challenge, and it takes time, it takes human resources, it takes
monetary resources to have it success.
So we have seen Nansen growing and developing
interacting their analytics tools into a multi-chain scheme.
So I wanted to ask to the Nansen team,
what are the challenge and opportunities
that you see when we're building AI power applications?
So what are the...
If you are talking to a new founder interested to get involved inside of AI,
what will you tell to them?
What are the opportunities that they are missing for not participating on AI?
Yes, so there are several questions.
I was reflecting on what Cameron was saying because in the end, what NIR is trying to do is really
integrate AI to an on-chain ecosystem.
And there are this challenge that are actually interesting, would be interested to hear more about
the signing issues, the safety issues and so on.
For us, it's a little bit different.
We are actually an off-chain product in a sense that we
We deliver our value of chain via web API or an API.
But the data primarily process is collected on chain via nodes.
So basically for, I think in terms, I can take our perspective, for example, first and then take a more general perspective.
But we've realized that more and more, especially in the past one year or so, that
that's not using AI, basically, is just on the one hand, losing on speed,
on the other hand, losing on insights.
And then also your customers are transforming the way of consuming data.
and they expect that you actually have this data delivered for them.
So what do I mean?
So I give you three examples.
So on speeds, for example,
we used to have quite a lot of manual processes of labeling data,
meaning attaching an entity,
is it a phone, is it a Dex, et cetera to wallet address.
And now we've pretty much automated that thanks to AI.
by training the models, giving it access to the data we usually look at,
giving it feedback, creating a database of feedback loop.
And this is right now running mostly thanks to AI.
So we are more efficient. We cover more labels.
So that's the efficiency and productivity part.
There is a part also around the user, I would say, that is also important.
More and more users are expecting to really not necessarily connect to a web API and look at a few charts,
but maybe have a prompt and be able to ask a question and then an agent goes through the different data and just
knows what to answer and what work stream the client is interested in and just go and grab the data.
So there's also a way of consuming data that is changing quite fast.
And we also have to adapt to that.
So I think from to abstract away to a more generic AI use case for founders,
I mean, you kind of need it because
Otherwise, you left behind in terms of efficiency and especially when dealing with data.
And also because it's going to be the new way to consume actually data and to come back to Cameron Point with agents interacting with each other.
going forward, it's probably pretty considerable to have basically an AI agent of a hedge fund or kind of a trading client or any kind of client interacting with our AI agent and nonsense and just getting the data they need, getting the insight they need.
Okay.
Interesting, and I probably will take this,
taking a little on this point.
Being honest, and I have seen many of the agents
and AI integration that we have for blockchain and AI,
Most of them looks like, let's say it like very, very, maybe it looks like many interactions of what we have on AI and blockchain is just limited to making, to putting a prompt, a prompt, putting a perm inside of an agent.
And this agent's making a transfer inside of blockchain of any tokens like sending.
its amount of tokens from account A to account B,
and that's like what we have seen so far.
This was our low-hanging fruit that we have on the ecosystem,
but I know that the new AI team and the announcement
AI team have a lot of experience on whether the needs
that are emerging on the market based on the capabilities
that we have on AI.
So AB, can you tell us,
based on all the experience that you have seen so far,
what are the most innovative use cases
that you are seeing emerging in the Web 3 spaces,
or those that are pending to be developed,
but will be a round-breaking opportunity for the people that start working on this.
So what are like these...
What are the missing pieces on the AI ecosystem that we can start working right now on this moment?
Who's that question for?
For the Nansen team.
Okay. I guess the most innovative is really the kind of the project that are trying to solve really big pain points for Web3.
One big pain point is really, I think it was mentioned at the beginning of the discussion is it's so difficult to use it.
There are like tons of wallets, some of different chains.
The interfaces are really, we talked like since years about simplicity of interfaces.
It's not really really achieved yet.
So what NIR is doing, for example, in terms of just attempting to basically make the user journey a little bit
easier in terms of cross-chain navigation, in terms of having agent doing stuff for you,
is really something we haven't really seen actually yet in terms of being operationalized and being practical.
So to me, that's really something that is really innovative and will address really a big pain point.
And then I'm just thinking also about anything to do with efficient smart contract operation,
solving the, that's why I wanted to go back to this point about safety.
Can we just make the processes on the blockchain also automated via agents and safe at the same time?
Because then that would be really a huge driver of adoption.
I'm just thinking, for example, what's happening in the
In the US at the moment, we're really going to about to see stable coins being adopted as really official and valid means of payment, especially for cross-border transfers.
It's already useful because you can transfer cheaper and faster across countries, but it would be even more useful if you add kind of an agent component to it, and you can do that in a safe.
in a safe way. So yeah, I would say solving the safety issue with agents and also really
lifting the pain point of interface, of interacting with this blockchain without having the
painful experience of different interfaces.
Okay, cool.
Cameron, do you have an opinion also on this?
I think that you will also have, like, a lot of ideas.
And what are these innovative use cases that are emerging on the markets?
So, one, I just want to adjust what she said is, like,
Definitely the role of agents is to make things simpler, but the security concerns are very real.
AI, even the best LLOMs out there, hallucinate all the time.
You don't want your model to hallucinate with your money.
That's a big problem.
So we absolutely – this is beyond crypto, by the way.
And this is also kind of where things get confusing because –
Everything about web AI changing interfaces for users also extends beyond crypto.
I would really say, like, the next evolution of technology generally is going to be more
voice driven and more AI powered.
And I don't think we're going to be clicking as much as we are today because these AI,
these LLMs or agents are pretty much going to know what we want in a as,
as long as the memory layers is verifiably private. And so right now, like, there's all these AI
models really don't, like some of them do store context of like the user, but a lot of them don't.
And so every time that you want to go, you know, ask an agent to rebalance your portfolio, it
it needs to know how you generally like your portfolio balance, what level of risk you're willing
to take. And so a lot of players have not solved that problem yet, partially because of the lack
of sort of privacy guarantees at the memory level. There are a few teams that are focused directly on this.
The ones that sort of sparked my interest the most are X trace. They're actually on one of our,
like we work very close together. And then Nillian is another one. So I
I'd say on the crypto has really been a great mechanism to incentivize
more research in applied cryptography.
So I would actually say one of the most valuable things that crypto has done
is fund a lot of ZK research and other applied crypto research.
So I would say like that is a very important incentive.
And the other thing is taking a bit of a step back to say like, well, what is the real point of
crypto? And up to now, you know, it's been for payments, in my opinion, payments,
and aligning incentives around how things, how money is allocated.
Because really, the crypto first utility was Bitcoin for payments, and the second was ICOs.
How do you crowdsource money to fund certain projects?
I think that that second use case of money coordination is very valuable for an industry like AI that does not have very clear and sustainable money.
monetary like funding mechanisms like it costs dozens of millions of dollars to train one of these models only for it to be out-competed by an open source competitor a month later like that business model is terrible so it also has a very big issue around content
attestation, pretty much saying that this data was used to train this model.
The New York Times is still in a lawsuit with Open AI for potential misuse of their
copyrighted material or their intellectual property.
And so I do think crypto is going to serve a very important role in rewarding those initial
contributors for model training.
And this is really important, especially in the world where AI is likely going to end up taking a lot of people's jobs.
We need a way to ensure people are compensated for their contributions for those models.
So I think the real use case of crypto and AI is around, again, payments and agent-to-agent payments and coordinating capital to democratize ownership.
of these large models.
In that world, I'm very interested in primal intellect
and what they're doing for to train models in a distributed way.
My only challenge with that is the close source model companies, Open AI, Anthropic, etc., they move much faster and don't have as much like, yeah, they just move much faster.
And it's like these decentralized train models are sort of two generations behind the cutting edge.
And so they definitely need to compete on a performance level to actually win mine share in the AI land.
So I think crypto is going to be used for the same mechanisms that has always been used, payments and coordinated capital.
Yes, I think that I'm not really properly.
It's because we have a different frequency on how our things being developed.
I mean, I will make...
I will compare, like for example, we have seen, there was a timing before Dipsick and after Dipsick,
taking all the politics outside of what this involves, but making an open source project
that allows you to run an LLM on your own machine without the need to be paying to private companies like the
the idea that we can found inside of blockchain and Web 3 for the development of AI.
And I mean making these kind of things probably takes longer.
It is harder, but once you achieve the milestone that you're looking for,
it always makes a shame on the paradigm of how are things being done so far.
Because, yeah, in the private side, things can be,
always a product. It's that. Always you have a product that people are interested to buy.
But in the other side, like the way on how are you handling your TESHA stack, having options on what you can be using at the moment.
So that's the point that we can have here on Web 3. And well, I think the, well, I mentioned that on 2025.
Surely there will be surprises on the...
next stage evolutions that we have.
Let's hope that this can happen this way,
and we can have the perfect alignment between Web3 and AI
to make this possible.
Cameron, I also want to discuss, and probably
Also making a wink in other side is a house of a stake is happening right now. I know that you are a delegate because I'm a delegate of the fierce delegates that are inside of house of stake
Of state for those of you that are new inside of the near ecosystem is
the new governance
hope that is working to distribute the new treasury to the projects that supports the new ecosystem.
So, probably you can tell us that how do you see that AI can assist to make governance better inside of the new protocol,
inside of any Web3 protocol, how AI can take action on this side?
Are you talking or are you mute?
People, can you hear me?
We can hear you, I think camera is unmuted, but maybe he has an issue with the sound?
Yeah, seems that we have an issue with camera.
Yeah, we don't hear you. Somebody else can hear Camero.
No, I think that you will need to, yeah, to go out and return to the spaces.
But you know, always these kind of things happen.
Yeah, do you have a comment about this?
How do you think that AI can assist the centralized governance
so that we can have better governance mechanism
inside of the different doubts that we have?
That's a tricky topic because in the end,
governance is really about people, psychology and incentive in a way.
So you could replace everybody with an agent, I guess.
and define tasks and define responsibilities.
I'm just thinking we did a sort of mini-dow at Nansen.
It was a mini organization to deal with customer requests and customer complaints.
And what's really cool with agents is that you can give them rules and then you can have like an art chart and you can give access to certain resources to certain agents and then they can interact with each other.
You could have, for example, the supervisor agent and then you have the agent that collects the data, the agent that cleans the data, et cetera, et cetera.
So I'm just thinking with DAOs, you could have to, you could kind of have the same in a way.
And define the task and the limits and the data of each of the agent.
That could be, that could be one model.
If you let it run, though, we go back into safety, safety issues probably.
And I don't know how the model evolves.
The second more practical way could be to just have AI a bit of an assistance to the DAO, making sure that processes are enforced, for example, having a bit of automation and having the DAO being more productive, more reliant. So that could be the second.
the second part. I quite kind of like the idea of the agent dial, but I could see it running
wild after a while. Yeah. I think Cameron is a listener. He's not the host anymore. So my
issue. Yeah. Well, there is no surprise that spaces is having issues. It's something like
very common, sadly, but it's very common.
So probably we will need to look for the next platform
that can't take lead on having space.
But yeah, it's really common.
So I think we don't sure worry about having issues
on the connectivity of some of the speakers.
But yeah, I think that we can follow on this conversation.
Well, probably I want to make a focus on what Nansen is creating.
Nansen is, well, right now they have a platform where they can provide
on-chain analytics for investors, and they are taking advantage of AI technologies to make this.
So that's something that right now we have inside of Nansen.
Can you tell us how AI is making improvement inside of the analytics that we have inside
of the blockchain ecosystem?
What is the key highlights that you have for the ecosystem?
Yes, so the first highlight is really,
instead of using traditional clustering models,
for labels, you can use also AI models,
and they learn from growing databases and feedback.
So that's first, that's the main improvement,
because a bread and butter is in the end just making sense
of the on-chain data that is quite anonymous,
and figure out, okay, is that a text swap, is it
Is it like, who is this whale, who is buying this Trump token?
So this kind of questions, you can use AI to help you build patterns on that and be faster.
You still need a bit of oversight and a bit of validation from a human because of all the positives and so on,
but it's faster for AI to do the job.
first. And then what we are really, really trying to do at the moment is improve the user experience
and maybe bring more off-chain data, so not only on chain to the user and having the
user being able to build its own question. For example, should I invest in that token? And then
the AI agent goes to different platforms. It goes, okay, how concentrated is that token?
in ancient wallets, so that's the ancient part.
And in terms of the price, do we see any price signals, like a simple stuff like momentum?
Then you go to social media data and then, okay, is there any news flow or any sentiment change around that token?
So really gathering parsing data and capturing it to the user prompt is really the area of improvement for us right now and what we're working on.
Oops, I was speaking.
Awesome.
Yeah, I think that it's worth to mention that Nansen is providing all these analytics to the people that is interested to participate.
Also, these spaces is possible because Nansen is also supporting the decentralization of the NIR protocol network
by participating on the enterprise program that we have inside of NIR.
We know that we are talking about AI, but probably you can tell us a bit of how are you supporting or how is your journey participating as a validator inside of NIR that you can tell us more a bit about your validator.
Amy?
Yes, a little bit because I have to say I'm on the research team and not the validator team,
but NIR is very important for us and it was one of our first partners and business that
we're really, really growing basically.
And we think that has a lot of potential.
Yeah, so that's all I can say, but the partnership is very important to us.
And we're very excited to see us all the developments.
I think it's one of the only protocol that is really working on different elements of the AI value chain.
So we are so excited to continue the partnership and see how the chain evolves from the DAP level to the infrastructure level.
Yes, I mean indeed.
The main topic on this is to mention that Nansen is part of the validator set inside of NIR protocols.
So anyone interested to support also the Nansen road inside of the new ecosystem.
You can go directly to your wallet and see that they are there for allowing you to stake your NIR tokens in case that you are interested to do so.
So thank you for that and thank you for being part of the near network.
Cameron, are you back?
No, I think I don't hear you.
I don't hear you at all.
AB, do you hear Camero?
No, I see that you tried to unmute, but I can't hear anything.
Maybe you were taken over by an agent's camera.
Yeah, yeah.
I'm also saying that the microphone is open, but we don't hear anything.
Okay, yeah, well, let's hope that we can solve the issue before we end the hour in case that we don't.
Well, it was still like a really interesting point of view shared by camera on all of these topics that we have been covered.
So maybe probably we are on the last track of our...
of our spaces.
So I always like to ask for an ending statement.
What are we missing to say to the audience?
What are we missing to mention on these spaces?
Are you there?
Yeah, just thinking,
I guess for me, okay, I'm a bit more on the investment part.
So I'm just always thinking where is the value flowing.
It's really interesting how we touched upon that a little bit at the beginning.
Where is the value occurring?
Is it really, we know it's not like the models being,
look at kind of the model development parts and the part that is really energy intensive.
So it has to be somehow the application or the DAP level or the any way the the part that kind of benefits from the AI model being built.
So for me, this is really the main question, both outside crypto and inside crypto when thinking about AI.
Yeah, so I think there is like from the perspective of the of the listeners, this is I think the important part.
And it would be interesting to see when you have a protocol launching and actually genuine adoption, genuine fees being generated from it.
And I do think for that you need to solve some of the pain points of the users and bring the value somehow.
And we go back to what we talked about at the beginning.
I think on the interface part, the safety and saving time part.
So yeah, that's my final word.
Okay.
Well, awesome.
So thank you very much for being good with both.
Can you guys hear me now?
We can.
Yes, we can.
Oh, right.
Twitter sucks.
Sorry.
This is terrible.
Yeah, yeah.
In fact, because I'm still, for me, your account is still a listener.
He's an speaker.
I'm not, but you can hear me, so.
Yeah, we can.
Okay.
So, yeah, sorry about that.
So I just quickly can kind of answer your initial question about how to stake, if interesting.
I do think that governance is inherently biased.
People are inherently biased.
If you don't speak the same language as somebody else,
you can't communicate with them as much.
And so you inherently trust proposals that come through,
let's say, a DAO,
if it's well written.
So how can we actually, in your native language,
the proposal might be amazing,
but if you can't understand it, you're biased against it.
And so I do think AI is going to play a very important role
in addressing bias in human decision making.
Obviously, that does require the models to be a bit more fine-tuned
on exactly the context of which they're trying to identify bias.
And so that's where I think that having some sort of AI model or a series of agents for the House of State is going to be really important.
For those, again, who might not be super familiar, this is an experiment to decentralize the near foundation to allocate capital.
And so, yeah, people need to be open-minded on the different types of approvals that come in.
So I do think that's going to be a very important piece that AI can help address.
Yeah.
And then the other one is, you know, decision-making is complex, like summarizing all the information, gathering context, very similar to what Nason does on chain.
We need to solve for a bit more of an off-chain human coordination piece.
So, like, if I don't know the context of a certain, of the history of, uh,
between the person applying for a grant or an investment and the person reviewing it,
and AI might be able to scrape the web and help identify where this relationship might be,
it might be a conflict of interest, let's say.
And so I do think that AI agents and fine-tuned models are really going to help governors,
like people that actually decide on where money is flowing, make better decisions.
So I want to flag how important that is, but at the same time, we can't be over-reliant on this stuff.
These models hallucinate all the time.
And so people still need to do their own research.
I think we're still in the sort of like,
beta or even like the alpha version of all of this where we should not be relying on it fully
but we can treat it a bit more as a as a tool to help us make decisions not as the mechanism
to make decisions so I think just to quickly answer that that's my thought and I agree that
surely we will have a long conversation on
Can AI take decision for us or still we will need to put our own human feeling inside of the decisions that we are guiding through AI?
That will be like...
a philosophical conversation that we should have around AI,
but still with points that you cover for what governance can look like for AI.
We were also saying that we are in the last track of these Twitter spaces.
So, Cameron, something that we are missing to say inside of these spaces.
Yeah, we recently came out with a paper called Decentralized Confidential Machine Learning, DCML.
Ilya presented this paper at Navidia's GTC conference last week,
and if you're at all interested in better understanding how the combination of verifiably private model hosting
can be used for properly paying people for their...
I guess, contributions to agent networks or like model fine-tuning.
It's a great place to learn.
I'll drop the paper in this thread because this is sort of our take on how we can actually
make the traditional AI business industry more sustainable to address that concern I mentioned before
around these business models of these large AI labs not being sustainable.
We do think that this is a prerequisite for AI also to be user-owned.
We're very skeptical of one company being able to sort of decide on who gets super
intelligence and who does not.
We believe it should all be open source, but as we all know, open source is normally wins,
but it's the businesses that really win or those that monetize open source, which is
But we need to make sure that those businesses provide back to the economy in the ecosystem.
And so we think that this decentralized confidential machine learning is the way to do that.
So I'll drop the paper here in the thread and highly suggest you look into that.
And also, if you're building useful agents, you know, agents that have very clear benchmarks of success,
I love to chat because at the end of the day, what Near AI is building is sort of a white-labeled super app,
making every app into a we chat of sorts through a network of verifiable and private agents.
So that's like big vision, obviously step by step.
So I can sort of drop the near AI roadmap once it's public.
So stay tuned.
Fantastic, thank you, Cameron, and thank you, Nancy, for being part of these spaces.
And I mean, really really appreciated all the ideas that you have shared with us, all the point of views,
and all the experience shared with the audience and the community.
So we are still going to have more spaces to talk with more validators that are supporting the Internet.
In this specifically, this one, we talk about AI on here and also AI on the blockchain ecosystem.
So we...
We would like to know that someone gets inspired by hearing all the ideas that we have.
So thank you really much.
I think this is all on our side.
We still have also a couple of AI initiatives.
We have an AI round happening right now in Metapool.
It closes the submissions tomorrow.
So if somebody wants to take a look, it is on our official site.
You can go to metapole.appap.
And find forward our guidelines for applying for the rent.
So thank you, everybody.
Thank you for joining.
And thank you, Cameron.
And thank you, Nancy, for being part of these spaces.
See you soon.
Thank you for having me.
Thank you.
Host
Meta Pool
from
Meta Pool ETH
