Projects
![]()
No projects found
Investors
New
DAO Treasuries
DATs
Lend / Borrow APRs
Chains
Categories
Spaces
🤖 Agents Unplugged: Your Bi-Weekly AI Deep Dive #6
Recorded: April 2, 2025 Duration: 1:20:49Space Recording
Short Summary
The discussion covers several key developments in the crypto and AI space, including the launch of new projects like Swarm Network's multi-agent framework and Oasis's privacy-focused L1. PAL AI's SwingX trading agent and its partnership with IBM highlight strategic growth and innovation. The conversation also touches on trends such as the increasing centralization of blockchain networks, the role of ETH in supporting DeFi and AI, and the broader impact of AI on automation and job markets.
Full Transcription
Thank you. Thank you. Thank you. so
all righty i think this might be everyone we have a full house today um i might be everyone.
We have a full house today.
I might be missing somebody.
I think I was counting seven.
Either way, let's get it started.
Welcome, everybody, to the sixth edition of our Agents Unplugged, your biweekly AI and AI agent deep dive all the way into the deep abyss of what is
happening in the space. Bullish as always, a lot of things are happening. And today is a bit of a
special day because we have a huge lineup of speakers. So my name is Yannick. I am the founder and CEO of Swarm Network. And I'm not
going to talk too much about Swarm. And I would suggest everybody that is about to do introductions
to just take between 20 and 30 seconds to introduce yourself and your project very briefly.
So I'll start Swarm Network, multi-agent collaboration framework, and a truth protocol that allows arbitrary data to be brought on chain, verified in a grandiose manner, and then be used by our
agents. We have a no-code tool that allows you to prompt agents up to 100 in a cluster, then add
them to a swarm and launch a unique asset for that swarm. That is about me.
I will pass it on to my
fellow co-host first, Marco
from Oasis.
As always, very pleased to have you, Marco.
But for the guys that are
new here, let's just do a quick intro.
Yeah, pleasure to be here again.
And seems like your beef
with Twitter is over. I did my post
last, actually, two weeks ago so
but yeah i'm glad staying in the meetings and not being kicked out to be fair it is from the
swarm account so i'm not sure if your personal one is uh fine with the algorithm uh i'm marco
i'm head of ai at oasis so everyone listening in should by now know Oasis rather well. We have built a privacy-focused L1.
And currently, we're working 24-7 pretty much on a T cloud.
So meaning any application that you have, and this is, of course, also agents,
can be hosted in a trust execution environment.
And through this, you get some guarantees to hopefully achieve trustless agents.
Cool. Thank you very much. What about, let's pass it on to Chris.
I guess Chris is joining us from the Powell AI account or Columbus. Oh no, that's yeah.
There you go. Chris, go ahead. Yeah. Hey guys, how's it going? I'm Chris. I'm on the core team for Powell AI. I'm sure everyone's heard of pal we're a cutting-edge
platform that integrates advanced artificial intelligence to enhance interactions for both
businesses and individuals we offer a large range of services you know we have pal as a service
customizable ai bots tools for crypto research analysis and in real life tools for, you know, whatever
field.
We do things like data visualization, revenue sharing through staking, secure data handling,
and we also have been focusing a lot in AI agents as well as AI trading bots and stuff
like that recently.
So I will pass it on to i guess change gbt
hey guys gm i hope you can hear me you never know with the x so i'm sharon uh broadcasting
from amsterdam cmo at change gt. Really briefly, we have three verticals,
AI company, we're developing AI application based on blockchain technology.
We have two launchpads, I will say a very successful one, if I may,
and an investment arm.
Happy to be here and share some insights and learn from the other colleagues.
Who is next?
Yes, let's pass it on to Freddy.
Yo, GM, GM, thanks for having me.
Thanks Marco for inviting.
Real pleasure to be here.
Yeah, I'm the co-founder of Nuvelati AI.
This is the tool that we made to optimize our portfolio.
Basically after many years in the trenches and
building product, we decided that it was suboptimal to manually rebalance and go through, you know,
scatter balance here and there, manage perps on Hyperliquid and go on a spreadsheet to journal.
So yeah, we just decided we couldn't find anything on the market. And with my co-founder, Mark Remark, we basically built it.
So we are an intersection of crypto and productivity apps, basically.
So you go on Nuvolari and you optimize everything for your portfolio
and reclaim time over no meaningful blockchain activities.
So you can actually have fun on a Sunday night
instead of checking your Excel spreadsheet.
Awesome.
And then we have two other speakers, Thomas and Paulius, both joining in from the Whitebridge account.
Guys, go ahead.
Introduce yourselves.
Good morning, guys.
Happy to be here. And yeah, I'll pass to Thomas to tell a bit more about us.
Hey, hey. Have you ever thought, guys, who owns your data?
So if you would think for a moment, then you would understand what we're building.
So we're building agents to protect your digital reputation.
So that's in charge.
Awesome. Cool. Just a little note for everybody joining the space today. As we have more speakers
than usual, I would recommend everybody to use the raise your hand thingy when you want to speak.
raise your hand thingy when you want to speak. I can see, I can moderate better that way.
We're going to go over a few predefined topics. And if you want to join in, then just raise your
hand like I did just now. So thanks for the introduction, guys. We already have 660 people
tuned in right at this moment, which is great. So let's just get right into it and kick it off with our first kind of like discovery question, which is what is the most exciting thing in the AI agent space over the past few weeks that either you have done or you have seen somebody do?
seen somebody do or um yeah maybe you know like one of your friends or one of your colleagues or
or somebody that you know just happened to be on your timeline anything will do um yeah go ahead
chris you can go first awesome so yeah um for those of for those who don't follow pal and aren't
directly in our community um the most exciting development in the space
recently for us is the launch of our swing x ai trading agent so it's a fully autonomous agent
that intelligently tracks market trends and executes high probability trades across a
diversified portfolio of 16 blue chip assets right now so it adapts in real time to maximize gains in any market cycle,
whether it's a bull or a bear market.
So SwingX can place both long and short positions
and it analyzes new trade opportunities
and existing open positions every two hours.
Even if you manually close a position,
SwingX will continue to monitor the market for trading opportunities. If you open a new position, it might close it early for you.
And the subscription for our SwingX is over 80% off right now. It's $50 a month. And I would say
the conservative guarantee that we have from our back testing is 100% APY
for a full year and that's whether it's bull or bear market and I do have a pinned tweet right now
that shows the back testing so you can go try out the back testing yourself you can enter what
amount you would put in and at what time you would have put it in and you can see
the results that you would have had if you had been using our swing x ai trading agent so uh we've
actually had so that was a conservative estimate right we've had more bullish estimates that show
like 1500 apy for a full year and uh and that will be you know over 100 monthly but depending on the
market really what happens is it doesn't make it every month necessarily it wouldn't make that 100
percent but we might have you know five bullish months where it makes a ton of money and then you
know two kind of sideways months where it just holds on to your assets and makes a few shorts and longs
and then really maxes things out when it sees a short market or like a bear market. So you can,
like I said, click on my profile and feel free to join our chat or do the back testing or click on
the agent on the website and just get straight to it.
There's a YouTube video that I have in the post as well.
Other than that, we've also been expanding into Web2.
So we have 49 sales people.
We call them agents, but they are real-life people.
And they're working with us with IBM to sell PAL solutions to in real life Web2 companies.
And you can see on my wall that just a few days ago, we got a contract for 200,000 US dollars and somebody deposited 197,000 USDT. our IBM partnership where we have these 49 people working to sell our AI solutions to
Web2 companies because, you know, we all know AI is expanding really quickly and it's scaling.
And, you know, we like to think of ourselves as like the open AI for crypto and for the real
world that people can actually access and say, hey, could you build this custom solution for us?
And we're able to execute that and give them a fair quote. So we have right now, just for people
to look into, we have our MyMemes platform, we have PalX, and we also have Gambit Markets,
which has a sports betting AI agent that has over 20 years of data for every sport you could imagine.
And it will actually give you like a solid 8 out of 10 win rate, whether you're betting on single matches or parlays.
So that's an awesome platform that we built from scratch.
And a lot of people are coming over from Rollbit and other platforms,
people are coming over from rollbit and other platforms and they're they're making a killing
using our agent to actually beat the books and beat the bookies and place these sport beds so
that's uh enough said from me for now so i'll pass the mic on to someone else
damn that's super impressive i quickly checked it out can you maybe elaborate on the autonomous part
we have trading agents actually on here regularly.
We talk about it a lot.
But there's two main things generally in the discussion.
First, where does the intelligence come from?
Obviously, it's not going to be LLMs.
And the second one is how autonomous is it really?
And I saw in your post you're saying fully autonomous.
So it would be great to hear how you achieved this.
Yeah, so it is 100% fully autonomous so it would be great to hear how you achieve this yeah so it is 100
fully autonomous each person who subscribes has their own ai agent on hyperliquid and right now
it's live on hyperliquid we will have bybit and binance soon but so for example if you get an ai
agent and i have one it won't be placing the same longs and shorts
both agents will be placing different longs and different shorts and right now we do power it
through gbt40 we find that's the best model to use uh there are other models like we're model
agnostic we can plug in any model that we want and we've tried all the other ones
and GBT4O is by far giving us, you know, the most consistent and best results, I would say.
So it is fully autonomous and we will have, you know, like vaults, like you can have a vault on
Hyperliquid live that other people could deposit to for free and eventually that'll come along and that will
be one bot that's trading with all the assets in the vault but each different ai agent that
somebody subscribes to right now is capable of its own thoughts and its own trades whether they're
long or short you could open up a new position and, you know, the agent might look at that position and
say, okay, I think that this is going to go down and close that position for you. So it is fully
autonomous. And this is something we've been working on for, you know, like a year and a half
now. And we did have a model that we released last summer that was on Solana trading meme coins and
stuff. But we had the problem that a lot of
meme coins were rugging and whatnot so we've made this to be uh first of all it has 16 tokens which
you can see in my pin post so it will long and short positions on all these tokens just for an
example I put $450 on March 26th into my trading bot, and it has already done $30,000 in volume.
It's not up that much, to be perfectly honest.
You know, it's up $5 right now, but we have had a very back and forth market, you know.
So for it to do that much volume and not lose any money is pretty impressive, I would say.
And I have the
aggressiveness set to 100. We have different aggressiveness levels. So if you want it to be
more conservative and safer with your money, you can set it to 5% or 10% instead of 100%.
But yeah, it is completely autonomous. And we're going to have updates soon that you'll be able to
change those assets so
instead of it just being the 16 assets in my pin post you could change the basket to have just eight
assets or you could do you know just bitcoin and eth and basically anything on hyper liquid you'll
be able to add to your agent to have it trading with those assets and we do have more uh trading
models coming out you know we will have specific
long ones and specific short models but like i said swing x does both longs and short it's cross
margin and i see the most at you know 10 20 x and like i said for 30 000 worth of volume in five days, it hasn't made a fortune, you know, because obviously the
market's choppy. Once we find a bullish market or a bearish market, that's when it can really get
some consistent shorts and longs and then, you know, keep those positions coming, like keep on
shorting, knowing the market's going to go down but right now the agent is obviously having a
tough time because we have all these tariff announcements and the market's just bouncing
all over the place so it's having a tough time catching a firm trend so i'll see it placing you
know five or six open shorts and longs at the same time as it reads all the different tokens that are in that 16 asset basket and it's really trying to
like i said find a trend and we have the back testing so you can go and actually plug this in
for free and try it out and see what your model would have made if you had it live from january
to march for example and you can also tell that it's much better than buy and hold. So
if you had a buy and hold that did 30%, our agent would have done 100%, for example, or it might
have lost 30%. And we would have kept that at 0%. So you wouldn't have lost anything in those times.
So we have this as a proven model, and it's just our V1. So it's going to get a lot better than
this.
Very cool.
I mean, I have tons of questions, but I'll just have to do some research.
I think what's even more difficult to achieve with these trading agents is actually the voiding trading fees.
I mean, those really eat you up. So the agent with this little starting capital not being in the minus is already impressive.
little starting capital not being in the minuses already impressive like generally i think the last
time i checked on hyperliquid you need to get to like 25 million trading volume to reduce the fees
to not bother you too much and that's not that easily achieved even with high frequency trading
but yeah i think we have tons of other topics too yeah yeah definitely so I will get back to the question. Like, what is the most exciting development in the agent space over the past few weeks that you've seen? Maybe somebody else wants to pick up this question and enlighten us with some awesomeness.
nobody guys marco not even you yeah i was just waiting for someone else
yeah you asked for a raise the hand and i i don't know took me a while anyway i can go next
i want to talk on a bigger scope what's happening in the market uh not just where three go ahead script crypto people but uh there
is bigger shifts that happened uh last week uh i don't know if you know gemini 2.5 just released uh
last week and it's really good by the way if you didn't try it but i think the biggest uh
by the way, if you didn't try it.
But I think the biggest change
that they have 1 million token
in the historic context,
which is insane.
And this is really big for AI agents
because it's all about
when there is reasoning,
you need to understand what happened.
The whole flow, right?
And they're going to expand it to 2 million,
which is really going to change the whole market.
Who is going to lead, which LLMs we're going to use,
and building AI agents in general.
So as mentioned before, using ChetGP 4.5,
I think we will see a shift into new models
that are more suitable for AI agents.
And that's a huge difference, right?
The second thing happened also last week is Amazon.
They released their new SDK for AI agents.
And I think what's really interesting is their vision.
And their vision first is an
SDK so everybody can develop and it's um oriented towards consumer use so for example using Alexa
you can just ask Alexa to buy stuff and it will just automate the whole process end to end and i think this setting up the vision what can also happen
in crypto right so go there more sdks more secure protocols um consumer oriented not just bots
more sophisticated more democratized so i think those two shifts in uh in the context right
So I think those two shifts in the context, right, generating more reasoning, more autonomous agents, that's one.
And the second, Amazon is just sharing like what's going to be the near future.
It's already happening for consumer use.
And I think it's also going to reflect crypto because in the end, we are using Web2 technology. And we are very influenced from Web2.
So that's my take.
Yeah, I love that take.
And you know what is really important to note here is that, you know, like people like to make a distinction between Web3 and Web2 agents.
But actually, all agents, as a matter matter of fact are web2 products that just have
an extension into web3 right exactly you can just exactly you can you can just think of it as okay
you know like eventually like every single client uh on web3 or whatever somebody that wraps a smart
contract into some kind of like logic that is accessible either through api or just straight
through the smart contract you could probably package that that up into an MCP and it becomes an extension point in the
action space for these agents.
And indeed, Gemini, Gemini, Gemini 2.5 Pro, we got a grant for Swarm Network from Google.
We got a few hundred thousand dollars uh worth of credits and we've
been generously spending on that on the gemini 2.5 pro experimental version and so on and we could
already see like really that just the sheer speed of the models that google has produced it's just
like it just you cannot compare it to other models and then now also it hitting like crazy benchmarks you
know like even the the gpqa diamond um it's outperforming most of the reasoning models
um except for claude i believe and then um like ima 2024
even the live code bench it's, it's doing very well.
It's outperforming DeepSeq.
I don't think it's outperforming OpenAI and Grok Beta,
but yeah, at least like some of the benchmarks are just really great.
And then also just the speed of these models are immense.
And yeah, it's also a large multimodal context.
You know, indeed, like you said, it ships with 1 million token context windows, soon
expanding to 2 million.
So that means it's able to handle massive input spanning from text, audio images, and
videos.
You can start to transcribe video content, and you can even parse large
video files directly through the model, because it's multi model. And the context window is just
so huge, right. So just the capabilities of what is going to be possible using these models is
like, on the horizon of like, kind of exploding. But but obviously i have also my kind of like um
kind of like uh yeah arguments here against gemini and the google models as they are uh
relatively biased and usually i mean like we find it sometimes hard to get them to do what we want
especially when we're trying to verify information and you know it's like
it's kind of like boxed in compared to the other models marco i see you have your hand up like you want to pitch in let's go i was just thinking when thomas and polly should be screaming in their
office now because you guys mentioned early on like hey who owns your data and now we're talking
about amazon providing an sdk, context windows increased from Google.
And it's just like, yes, all this potential is on the horizon.
But at the same time, there's always also this massive privacy invasion
where we're just giving up more and more data to these big companies.
And there's no better solution currently out there.
And it's still crazy.
Oh, yes. these big companies and there's no better solution currently out there and it's still crazy oh yes before i pass it on to the boys at whitebridge um the recent explosion of gpt4o the image model where
i see people like updating their people posting their passports and like this guy to
create fake documents i just saw a whole thread on like some idiot uploading his uh his passport
image and then asking it to create visa images and it was just displaying his his personal
information i was like yo bro don't do this please go ahead guys i will pass it on to you
i just wanted to quickly say um another model that people really aren't aware of yet is IBM Watson X. So if you guys want to look into that and the pricing for that and plugging it in, they're really working on that. And it's progressing amazingly.
And since we are model agnostic, we've also looked into like some of these decentralized AI models.
Like you have some open source decentralized models, but they're just not really on the same level quite yet, I would say, as decentralized AI models.
But like you pointed out, you know, Gemini and other models like that, they will not answer certain questions, right?
Because they aren't allowed to.
It's not in their code to be able to.
And the open source decentralized models, you can just ask them anything.
So it's a completely wild contrast trying out one of those models
compared to these centralized models that everyone's using.
to these centralized models that everyone's using.
From our side, so of course,
there are a lot of other projects
who are building, I think, really good infrastructure
to separate what data is private
and what data is public,
like Nile and Havana.
From our side, if you just think for a moment like bubble
maps gives you like a view on what's happening in your wallet but think about digital you like
you're using i don't know whatsapp like probably on your phone there are like a hundred of apps
here if you would look in the x.com like some of you using pictures, some of them not pictures, somebody
have their stolen identity.
But none of us even considering how we manage that data, which is like spread all around
everywhere.
And that if you would think how many data breaches are happening every single day, so
you don't even aware if some app which you downloaded, I don't know, like six months
ago, data was breached
and your password and email is exposed so we think and what we're building that this area will become
extremely big extremely because the people finally will understand that their data is the biggest oil for the AI to learn about people. And you know, we
thought that it's easy. But honestly, then there is like
10s of people with a name six with the same name, surname, for
current AI LLMs to separate who is who they're not capable at
the moment. I mean, I don't know based on how they constructed,
how long it will take them to separate
because now they are giving like blended average
of Thomas Martunas like Popolis.
And like, they're gonna tell that it's developer and CEO
and they're gonna mix all the hobbies.
And because it hallucinates,
it gives you hallucinations as an output.
So we need to fix the data.
And by doing that, we need to think how to do it in a decentralized way
so that the small communities in each country would be managing those data points
or data access points, APIs, so we will not ending up in like four or five companies' hands, like passing all our data.
Yeah. You know, there's this podcast, The Sean Ryan Show, and he has a guy on,
this guy is named, I believe it's Ryan montgomery he's uh the number one well not i
don't know if he's the number one but he proclaims to be the number one ethical hacker who does like
a bunch of stuff among like he hunts like child predators and uh he actually catches like one
life on the podcast but what he also does is that he um he reveals some metrics and
i don't know how well grounded those metrics are where he gets them from but he created this um
this uh this this application that essentially connects or is able to perform search on the deep
web through like a bunch of databases that have been leaked i don't know if they're actually going out and buy those databases and so on but he literally whips out his laptop during
the show and he shows sean ryan um his personal information um which is just readily available
online it includes i believe his home address i includes i believe his password it it includes i
believe the the address of the uh the place that they are currently or at
that moment filming the podcast and he's just like blown away by it but that is a very real thing and
you know like we've spoken about this previously especially with marco it's like you know like
people don't care until they either are opened up to the almost guaranteed fact
that their data is living somewhere on the dark web
and some predator eventually might have their data
and do something with it.
And these risks are becoming like more real, right?
Because now all of a sudden I can train a model using a few of your pictures
and I can make it look sound and act like you.
I can set up, I can sim swap you, I can set up a video call with your parents
and I can get them to essentially like empty their bank account because you know like whatever i'm saying i'm
having this this uh this version of you saying that um i'm being held hostage while i'm on holiday
in thailand or some shit like that right like these things they sound super crazy they sound
like they're from a movie but these things are very very very very possible right now and it's
even more simple than that you know like you can just voice clone uh somebody and you have now all of these gaming applications that allow you to um to use
ai to clone your clone a voice and use that voice it's like a voice altering uh thingy that you can
just use while you're gaming online and so on so like these you know like this, like this is a real issue. And then the second thing is that, you know, they actually not just are aware or made aware of the fact that these are issues, but they actually experience something like in real life where their data has been leaked or something like that, you know.
And why is data really important to be kept private and to be kept secure?
I'll give you an example.
I just received an email from X on my personal account, my personal email address.
Supposedly it was from X and it was about the Swarm network account.
And it said that we had violated some kind of privacy policy and it gave me a case number and so on.
of like privacy policy and gave me like a case number and so on. And I obviously didn't click
on it because actually my personal account is not that email address is not the email address
that's connected to this, this account, this account has an email attached to it that nobody
knows. And it's like not obvious. And it's not accessible to neither me or anybody within the
team because we've experienced hacks before but you know to the
naked eye to like the unknowing and maybe relatively ignorant person that email looks
super legit and you're like oh shit i don't want my account to be banned or like get into issues
because like these these you know like copyright infringement things especially for example on
youtube they they hit pretty hard you know like they can just like kick you off the platform.
You know, this person is going to click on it.
They're going to log into their email.
They're going to use their email to log into what looks like a clone of Twitter.
And then they not just lose access to their email address, but they give away probably
pretty voluntarily the password that they use everywhere else.
And with that, you know, like, good luck.
Like stuff can happen.
And until that happens to you, you realize like, oh, shit, you know,
like even my email address shouldn't be out there because I become a target.
Right.
Like you become a target for real, which is a huge issue.
So, yeah, speaking about that um indeed big problems maybe uh freddie
we haven't heard you man maybe you could uh take it on yeah him i hear him i hear i know i've been
listening a lot i mean uh privacy is definitely one of the you know feel the more passionate about
just for from a personal productivity standpoint i try
you know my best to stay private back in the days there were a few uh apps before you know this whole
ai madness went viral to to stay to stay private you know we had first you know the password
managers that were completely offline then we had the VPN mania to you know give me
less you know data to the VPN provider I mean there was a huge candle so North VPN and so on
and so forth so encrypt is definitely one of the things I am most conscious about this one was one
of the reason why I actually joined the space in the first place after having a few problems with banking
and so on and so forth.
What I noticed, I mean, just to broaden my spectrum on the privacy side of things, is
really interesting to see how retail is reacting to this whole AI agent madness.
Because it's true that generally Web3 and Web2 things have been a bit of a silo.
We had crypto stuff and the rest of the world that was doing SaaS on the opposite side.
And for the first time, I see this sort of major interaction with developers that were
in Web 2 can finally hop in Web 3, but with use cases that generally are not biased by,
if you allow me to bring a random application that sometimes we tend
to see in crypto that i'm not saying necessarily bad but very narrow use case so i i see retail
really you know questioning in a different way with different tech stack which are the problem
that they might face we might consider them silly you know it could also be just a better ui but all in all is uh you know more
things coming into into the space just very broad scenario but vibe coding i know i'm gonna raise
one thing that marco uh you know loves and you know we had a few chat about that and i'm a big
enjoy of that but i'm not an enjoyer because you know i like this idea of having very niche apps and software
but that's a bit of behind you know the core philosophy of crypto you know having of course
you know massively accessible application but also having very narrow use case apps and that's
possible only people and retail have access to extremely powerful tools that allow them to develop
in a more streamlined way, something that no one else would have took care of doing.
I mean, just for the purpose of the conversation, I 99% of the time work
in the Volari, but free time I made like, you know, Tinder for Dune Analytics just
because I don't like scrolling to Dune.
So I just make a better UI for Dune Analytics because I don't like it.
It's too cluttered.
So I just select my preferences and I have the dashboard I want to check.
I want to check MegaEat, boom, I have that.
Very narrow use case.
Another topic we talked about with Mark when we started building Volare.
And I don't want to turn
into a shilling of that i'm just trying to broaden the scope of my of my speech and cut me off if it
gets too too long but was ai in general is really creating this sort of modular approach to software
you can really go live very quickly with one product that is a component of your bigger tech stack you know fully fledged
you raise 10 million you want to do something crazy but with the first 50k you can do something
else okay uh you you can narrow down the case and get feedback from that that's what a bit what we
did with uh so we started with a bookmark manager because i hate going through my bookmark i never
open them up i forget about them so i make a tool that connects the bookmark and because I hate going through my bookmark. I never opened them up. I forget about them. So I make a tool that connects the bookmark and summarize my Twitter bookmark,
but derives strategies that can execute on the Volari.
So I can click and execute directly the yield strategies on, I don't know,
Sonic or I'm a big hyperliquid enjoyer for the past year and a half.
I can do whatever on Hyper-EVM directly from there.
Now is accessible access code, whatever, end of the show. last year and a half, I can do whatever on Hyper-EVM directly from there.
Now is accessible, access code, whatever, end of the show.
But this is just an approach of modular product.
Then we're going to move on LP notifications because I don't like checking LP out of print.
This is possible because AI is really streamlining the code base and the retail can give you
a feedback very, very fast. And yeah, I mean, I know we started from privacy.
My last point on privacy is that I'm a bit worried about MPC, MCP's privacy,
mostly when it comes to on-chain MPC and the fact that you have to give the private key.
I think that was the base one.
Yeah, that's my last link to the initial topic.
Sorry for the detour. Hope
raised some interesting point from your side. Yeah, no, it's not a detour at all. It's actually
really cool. And I wanted to ask you a follow-up question to take you a little bit deeper, because
this allows us to kind of like segue into like a space that we hadn't pre-planned, but you're really talking about how the kind of like data
and sentiment data that is happening on X
is able to feed into your engine essentially
and based off that data, you can then execute trades
or execute trading strategies using your agent, correct?
Yeah, I mean, my thesis is a bit different from the general agents.
And we don't fully trust LLM.
As I said, we started making this product for our own portfolios.
And if I have to make 100k, 200k, 200k transactions,
I cannot afford to lose it. I mean, I'm not billionaire.
Okay.
So I really have to make sure that the transaction goes through the way.
So we decided to go shortcut first and we powered it inside.
So I optimize the data that you receive, but I'm the ultimate decision maker on what
I have to do.
Most of this is very clear when I think colombo uh raised the point before you know
on perps uh perhaps really need a lot of in my opinion uh consideration on the position you can
open so they cannot really be fully automated other than delta neutral strategies so what we do
we just aggregate information where you get the information you get it on twitter we enrich that
data and then we allow you to execute but all if you want and after you assess the information, you get it on Twitter. We enrich that data and then we allow you to execute, but only if you want and after
you assess the type of information.
So we just optimize the bummer of open up the Twitter bookmark, collecting more information
and of course we are chain abstracted.
So there are a few perks, but yeah, we just focus on that part.
I don't want to be delegate my full portfolio to an agent just because I don't trust
and I believe very few people also in this space have deposited 200 300 1 million into an agent to
trade for them but that's my take go ahead Chris yeah so one thing I wanted to bring up is when you're, you know, doing pre-sales and like venture capitalist opportunities, you have to fill out a lot of stuff, just like a KYC for an exchange where you're giving your name, your address, your license.
You know, sometimes they even ask your passport too.
yeah your social insurance number like crazy things so if one of those places has the data
Yeah, your social insurance number, like crazy things.
leak then everything that you have is exposed somebody knows everything about you and all it
takes is for you know like one pre-sale platform that you just signed up for just so you could get
in on one pre-sale to have like a small data leak and everybody's information all of it is out there for anyone to
see so i'm always super careful about uh about where i like you know give my information to but
just recently um i did a a seed sale for red belly network which is an rwa project and uh to claim
your tokens on their blockchain and do a transaction just to even
bridge them to eth you have to do a full kyc and even give your social insurance number so it was
like wow like i had to think about it for a few days man i was just like this is crazy like but
uh in the end you know i talked to them and talked to their team and they sounded pretty safe and pretty confident in their security.
So I just I did it anyways.
And then I just staked my tokens because it does look pretty bullish.
I'm super bullish on real world assets, especially real world assets and AI.
And but, yeah, I just wanted to bring that up.
And I know Chain GPT has like a pre-sale platform.
So you guys might even have to get a lot of this KYC information from people.
I haven't used it, so I'm not sure if it's based on that or not.
But it's just, it's crazy how much information they can get.
And for me, the craziest part is actually, we are aware of these data honeypots it's pretty
straightforward when you kind of have to show a picture of your passport you have to scan your
face you have to enter all of this personal information that's pretty much in your face
like hey you're giving up your data but what people are still kind of not realizing and this
is also what Yannick just mentioned this awareness of users isn't there is when utilizing any llm you are sharing
all of these prompts with a company and there is a company and they have built this massive profile
about you so they know tons of stuff and i mean you're sharing very personal things i am
yes i mean but google google facebook those were the big evils five to ten years ago everyone
knew like man they took lots of my data they built these shadow profiles so it was it was possible
to avoid like i could just stop using facebook i've never used instagram i don't have to use
tiktok i don't have to even utilize google i don't have to utilize G Suite. It was possible to circumvent or avoid it.
But with LLMs or AI, you can't. I mean, we had this discussion also a couple of weeks ago. Yes,
of course, you can run some models locally if you have the right hardware, but that's not
accessible at all to most users. So you kind of are forced to give up your data and you're paying
with your data
and there really is no better way currently because if you don't use this you lose massively like this is such a competitive advantage to be very close with a model so you kind of have to
i mean yeah freddy i i see you you had your hand up. Go ahead. Oh, yeah. Okay. I'm still not clear how to raise the lower down.
Yeah.
The thing that meresmerized me the most is, I mean, I think we went through the problem
of managing password.
I think most people now manage passwords correctly.
I really want to see how companies are going to handle, you know, chat GPT and stuff.
Problem.
I had a chat with a very good friend that works in a massive European
company a few days back, and a small team of guys reached out to the company
asking if they were interested in some sort of Chrome extension that would
prevent employees to upload sensitive data inside chat GPT.
Yeah, despite this company having thousands of employees, they didn't
consider that extension
as a valuable resource.
So they just prefer having employees uploading Excel spreadsheet with a full all in gas transaction
onto GPD with the name of all the buyers just to sort them out because they don't see the
problem.
So the question to me is really how we make the problem compelling the moment that we know that the enemy is basically
who's creating those tools you know gpt's and traffic and all the tools that we love because
we we love what the building they're making is making our life better but at the same time
the level of education that you need to properly use them is is way out of scope for for most of
us and most probably also uploaded something
that I shouldn't have uploaded now that I think about it.
So, I mean, really, how we make the problem compelling
to avoid major data leak.
Just as a small reminder, you know, just buying a ledger
a few years ago, if you bought a ledger,
most probably your address got leaked.
And we're talking about buying a crypto wallet.
And I mean, no one's safe at this point yeah exactly I want to remind everybody here of the promise
of of decentralization and the blockchain in general, right?
Like the promise here is that we can actually operate trustless, right?
Like we don't have to reveal each other's data identity to be able to transact and to exchange information with each other right
so I feel also a little bit like sometimes the space tends to forget that I just saw this
I just saw this post of like a very big data leak for example of I forgot the project, but they leaked like 170,000 something signups.
It was not, they weren't KYCing, but they were, you know, like just like name, address,
phone number, email, I believe.
But still, I mean, it's pretty wild, but that doesn't have to happen in Web3, you know,
like there are little occasions where we actually need to know each other's real identity and personal information, right?
Like most of the time what we should be aiming for and what we should be promoting is this, you know, like abstraction layer, which decentralization and blockchain technology really is.
So I don't need to know who you are.
understand that you are able to operate and that you are able to connect your wallet and that you
can you know like contribute some kind of value whether that's monetary or or otherwise but um
yeah that's really important you know like especially for us builders here that we remember
that and that we actually elevate that to the highest level you know what i mean go ahead chris yeah yeah so you just sparked a huge conflict
that i've been seeing a lot lately you know you've got all those eith bears that are posting like
eith is going to zero or whatever you know and if you look at satoshi nakamoto's original white paper about crypto. It's all about decentralization. Bitcoin
was created in the wake of the 2008 financial crisis, which is where the world was put into
a massive recession because there was too much debt and they were printing too much money and
giving it out to too many people in the short. So basically, that's what sparked the creation of Bitcoin and Bitcoin sparked
the creation of the whole space that all of us are in right now. And if you look at the top 10
tokens and you look at the Bitcoin white paper, the only other token that actually works as
something Satoshi Nakamoto would say that he trusts is eth because all the other tokens are completely
centralized you know like i'm not going to bash solana or anything like that but it's a perfect
example the chain has been turned off three or four different times and basically it's run on
supercomputers and nodes that you can't run yourself because you don't have enough compute power so solana can be
turned off you know it's all centralized all these chains that everyone is using now that aren't ether
layer twos are basically centralized chains that somebody has all the control of and can turn on
and turn directly off and basically if Satoshi was here today,
he would basically say that's exactly the opposite
of what he envisioned for the space.
So you see so much of crypto evolving
into centralized and institutionalized kind of space
where we've got all these stable coins
and BlackRock moving in now.
And it's great for crypto
to to spread the adoption and have everybody realize blockchain but eth is never going to
die because eth is the only pillar that we have other than bitcoin where you can actually start
up a node in your house uh using regular compute power as long as you have a few terabytes of memory for the blockchain,
and you can run that node and validate the blocks yourself.
And based on the fact that you can do that, that tells us ETH is decentralized.
We don't know who's running an Ethereum validator in their basement and who isn't.
So that's why I think ETH can't die.
It's the hub of DeFi, of AI, of real-world assets. ETH is what is keeping the space alive. And then all of the Layer 2s, their security is actually running off of ETH and the fact that it would take over 51% of the ETH validators to do any sort of hack on the network, right?
And to purchase that much ETH,
someone would drive the price of ETH over $100,000 per token
because I actually run ETH validators.
So it's just...
That would not be too bad, honestly, because it's tanking.
So it plays a 1-2-51 attack.
So the price goes up.
Yeah, yeah.
That would...
Yeah, no, that'd be amazing if eath was
at a hundred thousand dollars but i do think eventually one day it will happen and we look
at all these layer twos that have like 50 billion dollars on them like base they are all relying on
eath so if eath got too low i feel like all of those layer twos would chip in to make sure that eth you know survives
and is strong and prosperous because their security relies directly on ethereum so i just
wanted to to make that point and say you know i don't think eth is ever gonna die i think it's a
great uh buying opportunity for eth and a lot of the alts that are on ETH right now
Yeah, and you know like what the what the beauty about this is like a lot of people are hating on the foundation and so on
But the foundation essentially doesn't matter anymore, right? Because like the system will be able to run as it is running for the rest of eternity because the notes are
Like distributed right like nobody knows who's who's running what like you said
so, you know like crypto Twitter tends to be very dramatic and so on and they are shitting or distributed, right? Like nobody knows who's running what, like you said.
So, you know, like crypto Twitter tends to be very dramatic and so on, and they're shitting on Vitalik and so on.
But he's not the CEO of Ethereum.
He's like the main spokesperson.
He's one of the co-founders.
But ultimately, he doesn't have the veto.
He doesn't have the ability to truly change something and
even with the miss potential mismanagement or hypothetical mismanagement of you know like the
ethereum foundations their funds and so on it's like like who knows like who like you're shitting
on on on vitalik but he doesn't even have control over that probably you know like that is also just
decentralized and governance and so on so yeah that's how it's supposed to be.
TVL is on ETH end of the day so ETH is not going anywhere so
yeah and I see ChainGPT account had the hand up too. You want to speak?
Yeah, really briefly, going back to privacy.
I think that the main point that consumers, people don't care really about privacy that much.
And we might get into the point really, really soon that they will care a lot because the difference between brands like Google or Facebook and etc. to AI, people are afraid from AI.
thousand people, Elon Musk was one of them, signed this kind of petition on how to call it that we need to stop AI for six months.
Because the number one threat is existential threat to humanity.
If you unleash disintelligence without control, without ethics, without privacy, this is very dangerous.
And it's not in humans
and people are afraid
we are humans, we are afraid from the unknown
that there is faceless
so I think we are
witnessing the first
biggest use case for
blockchain maybe
privacy, security
you know, ZK proof
because this is
gonna just go public everywhere.
The threshold will come when AI will generate AI, and that's it.
It's out of our control.
Even now, I know if you guys read the Anthropic latest article about how AI thinks and trying to understand this reasoning.
It's really hard because it doesn't operate like a software.
It's being trained to think.
So we already started losing like, okay, can we control it really?
And elucidation and other factors without verification, without privacy, without trust,
this could destroy a society, democracy.
So I think this is the biggest use case for blockchain.
Of course, there is DeFi and cryptocurrencies.
But technology-wise, I don't think we can give the authority for governments or companies like Google,
yeah, yeah, we can trust you authority for governments or companies like Google.
Yeah, yeah, we can trust you.
You just, like you said about Amazon.
Yeah, Amazon, just let's use their SDK.
You will consume all my data.
Now you know everything.
And we'll put on my browser, my phone, and game over.
Then you don't even know.
It's like it's not humans anymore.
Who's going to control it?
Regulation.
So that's kind of my take on privacy and why it's important
soon enough our society is going to run by a code probably nobody wrote not a human and that's not fun to take five years or one one year but it's that's going there right society is a code you
you gave me an idea you gave me an inspiration what we're gonna end off with
in a few minutes is we're each going to take uh two minutes to give our three predictions
of how ai and ai agents are going to change um the world in any given time frame you can mention
the time frame that your prediction is made up of and i'm going to register these uh these
predictions and we're gonna probably meet at a later point and i'm going to register these uh these predictions and we're gonna
probably meet at a later point and i'm going to bring them up again but before we do that i just
want to i just want to tack on to um what you just said regarding the tracing of the thoughts of llms
and especially um anthropics research they they they unveiled these interpretability tools.
And what they can do is it's essentially like kind of like a microscope, I suppose you could call it.
So you can look into or you can track the you can trace the models, internal circuits, plans and kind of like real timetime conceptual thinking and there is just so much
to unpack there it's really really awesome i will just mention the like the bullets here because i
had prepared them originally for for some talking points but the first one is that there's multilingual
language of thought which sounds really complicated but but Cloud essentially processes concepts like small or opposite, similarly across English, French, Chinese, Korean, Japanese.
And this does suggest that there's like some kind of shared abstract representation layer.
So as the models go up, these cross-lingual features will be able to enable like transfer learning between languages really quickly.
And then we also can see that there's like some expectations that we had about how these
models plan have been contradicted because even when Claude is planning or writing rhymes
when Claude is planning or writing rhymes or generating rhymes, for example,
there's already kind of like a planning before the writing that happens.
So, for example, when generating the line, his hunger was like a starving rabbit,
it had already decided essentially on rhyming it with grab it.
So researchers could suppress or swap this
out to alter the ending dynamically which is very interesting um because it shows kind of like uh
almost like a human element to reasoning and to kind of like subconscious reasoning almost
um and then there was this mental math with parallel circuits, which basically means that Claude computes sums using parallel circuits. So one estimates the results, then there's another circuit that tries to predict or tries to get the last digit.
as how we would come to the conclusion of that math problem,
like one plus one is two.
It predicts what the last digit of an equation could be,
and then it kind of goes about it,
and then it sometimes doesn't even refer to the true logic,
but it will explain it to you like a human.
And that reveals a gap between the internal computation
and the verbal justification, which also brings us to another really key aspect, which is there was a lot of detection of unfaithful reasoning.
So a lot of times, Claude actually fabricated logical steps to fit a target answer.
steps to fit like a target answer. So it would probably know if you'd ask it again that the
logic that it derived or that it proceeded to tell you does not actually work. There's no rational
basis and heuristics wouldn't work if you would reanalyze the logic. But because you are suggesting
a certain kind of like hint towards the outcome outcome it would kind of try and build a logic
around that to try to satisfy your hint right so that it actually because like these models are
are supposed to optimize for a positive reward right so the positive reward is when the answer
that they give you is the correct answer and that's very abstracted by the way but
yeah it's very funny right and it's very interesting because this is also hinting towards a very human kind of like uh behavior and kind of like a human bias yeah marco you
want to say something it was the craziest part of me i mean i summarized it as the llms they feel
pressure they really want to answer correctly they They want to complete tasks. So they're kind of building up this pressure internally.
Of course, it doesn't make sense as a word here,
but they feel it and they really want to kind of satisfy you.
And that's the thing that kind of makes them human.
Like that's the point where they do mistakes.
That's where they hallucinate.
That's where you can jailbreak them.
And it's very similar to how we as humans behave,
right? Once the pressure is too high, we simply can't function anymore. Yeah. It was the craziest
part to me from this research paper. Yeah. Yes, indeed. And this also brings to light,
indeed, like these kinds of worries that we have, right? Once we, I mean, like a large language
model is definitely not the end-all AI solution
and there's going to be a lot more innovation
and there's probably going to be different ways
to create these large language models,
such as now you see these diffusion models, right?
Which are trained in a different manner,
which are doing very well.
But it brings to light a very reasonable concern, which indeed, ironically, now that XAI belongs to Elon Musk and that they're developing their models and that they're heavily promoting it and kind of like unleashing it into the wild, which is X.
But previously, he'd say that, you know, we probably need to stop developing.
But now it seems that because nobody is listening to him, he just went ahead and, you know, went balls to the wall,
which is very, very funny.
But yeah, it indeed, you know, it's concerning.
So we are at the end of the allotted time
that we had scheduled for this session.
Thank you, guys, everybody for joining.
And I would like to, you guys, one by one,
give me the three predictions regarding how the AI agent space is going to change the world.
And I will end off. So I would like to give it to the boys over at Whitebridge first.
bridge first so i think agents gonna be replacing one by one different verticals and feeding them
the data in certain vertical in certain role we're gonna be seeing more ai salesman ai business
developer ai marketer because you train one person and you resolve million times
so maybe one prediction we're gonna be having more and more super humans or super roles
moving forward in my view in a agent space
in my view in a agent space
cool um all right one prediction that's cool um marco what about you
yes um i have a few on the kind of short-term horizon i think we are quite close to these
always-on systems where whatever we kind of
need we just say it and there is something listening kind of like an alexa in your room but
hopefully not from centralized companies and it will just execute something i don't think
like agents are the solution for everything like i still think humans will be quicker
pressing just one button on my um phone to kind of check the weather is
quicker than asking an llm like hey can you tell me what's the weather in munich currently i mean
i can i'm quicker than this so there's different interfaces for this but these always-on systems
will really kind of speed everything up uh another one is and i had this thought yesterday
when kind of researching some topics that there will be this new role, head of agents.
And this is honestly going to be the new HR department
and sexiest position in the company.
So if I ever leave Oasis,
I will go towards something head of agents
or this will be the new founder title
because it's honestly like agents
will be replacing lots of your workforce.
So you need someone to manage all of those agents. And then then the last one kind of a bit of a longer time horizon and that's
when it gets a bit more depressing is that i think really we as humans we will need to start fighting
for our plan kind of position in this world because it is going to become dire once models advance
to become dire once models advance.
Looming, dooming, dooming and looming.
Sharon, what about you?
Very interesting question.
I see one scenario when there is an emergence of AI economy,
AI agent economy, which means they have their own ecosystem.
Language, infrastructure, transactions, building stuff, selling stuff, buying stuff, sharing data, intelligence.
So it's like a machine world, not necessarily a bad thing.
The second scenario is if we don't embed ethics, human ethics into the system and just unleash
them into our capitalistic environment, they will seek to gain power like any other entity
in the world.
There is rules that we are playing.
That's very dark future.
I don't think it's going to happen because in between now and this economy, something
will happen that will force us to stop and think, okay, where do we want to take it? Especially the
war between the U.S. and China. Every other week, you see a new model, cheaper, faster on premise, on machine, on your computer.
So this economy is going to be built.
The question is how fast and what about ethics or regulation or something is going to have
more constraints.
Very nice.
The third one, the third one really quick.
I see an emergence of AI-driven or AI-agent-driven companies
where 50% of the workforce is just AI agents.
That will be the winning companies of the near future this year, by the way,
not far away.
Yes, very nice, very nice.
I've written them all down.
Chris, handing it over to you uh yeah so i think um ai is going to revolutionize industries i think ai is set to transform various
sectors you know from health care to finance like in health care for instance ai could analyze
vast amounts of data in diagnostics personalized medicine even robotic surgeries
and finance ai algorithms like we already have with our swing x agent will enhance things like
trading fraud detection and provide personalized financial events so or advice, sorry. So like with the whole DeFi AI movement, I think that eventually portfolios, instead of being managed by people, even for hedge funds and stuff, I think eventually people will be able to trust AI with managing
their assets and doing trades for them, whether it's leverage trading or just buy and hold
strategies.
A second thing that I think AI is going to do is enhance automation and productivity.
So I think as AI continues to evolve, we can expect a huge increase in automation across all job sectors.
And this wouldn't just streamline operations, but it would also free up human workers to focus on more creative and strategic tasks.
And it may lead to job displacement in some areas for people, but it will also create new opportunities in AI development, maintenance,
oversight, you know, the people that are going to have to fix the bots so that they can continue
running the factories. And the third thing, I think AI is going to change personalized experiences
for everyone. I think it will enable hyper personalization and consumer experiences,
you know, from tailored marketing strategies
to customized content delivery. Businesses are going to be leveraging AI to understand
consumer behavior better and produce products and services that meet individual needs for everyone.
And this is going to enhance the way customer satisfaction and loyalty turns out and driving growth in
various industries. So people will be, you know, basically sparked to work with and more prone to
use platforms, even in things like social media that have this type of AI interface built into
them earlier. So really, I think AI is going to change the whole world, like everything.
And I think that in finance, too, just touching on my first point,
AI isn't just going to manage crypto trades.
It will manage stocks.
It will manage Forex markets.
It's going to manage bonds.
It's going to tell you where you can deposit to get the best APR and like a high interest savings account, whether that's a short term yield. So I really see AI just taking over pretty much everything in the future.
Awesome. Thank you very much.
Freddie, what about you?
Yeah, real quick.
I think they were raised very valid points.
So one doomed forecast I have without getting political is the role of Europe in this whole chessboard.
I'm not really sure how Europe is going to position itself compared to China and the US,
mostly when it comes to the job market.
A lot of people are going to get sidelined.
I believe this is a bit of an industrial revolution type situation where you either operate the machine or you back to the
field so um yeah i think a lot has to be done to make sure that people can you know smoothly
on board onto this and so yeah just true doom uh scenario is job market going
to zero mostly for entry-level jobs a A lot of education systems that we built so far,
mostly when it comes to universities, are not designed to implement these type of tools onto
the day-to-day activities of students. So I'm pretty worried that a lot of new graduates
are just going to hop in the job market with no clue and realize that a lovable can do their job,
or an NPC can do their job. And that was going to be very disappointing for a lot of students
to spend maybe 70K for a college degree.
But this is on the negative side.
On the positive side, I see definitely one scenario
in which we're going to have the first time solo builders building
100 million MRR products because they ultimately completed the flow of marketing,
you know, bike report and everything. So we'll not be surprised to see in the flow of marketing, you know, bug report and
everything. So I would not be surprised to see in the next five years, you know, the level CEO
of the future saying, yeah, 800 million MRR with this product would not be surprised. I see app
as ecosystem. So we're going to see AI really pushing the boundaries of what in-developed
developers and small team can do and expand
onto the application. And onto the final point, which is the most positive,
I believe it's going to be a very egalitarian market. The steep increase of apps is going to
create so many different competing software that developers are going to be forced to really optimize for the
end users and basically being forced to find product market fit immediately to justify
committing to that product and this all in all i believe in the self-regulated market and most
when it comes to software this is even more more clear and yeah i i believe is all no positive
more competition means the software gets better and better immediately.
We can see already in the last five years,
the level of onboarding that you had five years ago
and the level of onboarding you have right now.
And now with AI and standardization on a few flow,
it's gonna be even easier for entry level developer
to build cool stuff.
And yeah, I mean, although I'm positive,
just a bit bearish on Europe and job market,
but I don't know that it's not good. Beautiful, beautiful. I mean, although I'm positive, just a bit bearish on Europe and job market.
But I don't know, that is good.
Beautiful, beautiful.
My first prediction actually ties in and relates to your prediction regarding the egalitarian markets, which is that I believe that there will be a fundamental shift in the way we build and invest in companies. We've kind of touched on it earlier
a little bit as well in the conversation, but essentially what I believe is that
the traditional way of building software is going to fundamentally change, and with that we are also
probably going to access the information age
where software is purely built for the information highway. I think there's going to be a lot of,
I like internet of things, but also deep in kind of like integrations into the environment space
of clusters and swarms and agents. And they will be able to really coming back to marco's point i suppose
tie in and always be on and really consume information that is coming directly from our
environment and that's going to fundamentally change the game so yes so essentially first
change in investment strategy for vcs and and also change in strategy for builders that is going to cause us to
have a more egalitarian market which is more optimized and will also force us into a direction
where we will exit the pure software information age. The second prediction would be that within
the next two years, all middleware is essentially swarm or cluster or agent run.
That means that every single application that you interact with has somehow integrated a middleware layer that is purely consisting of agents and clusters and swarms.
That could be a business that you interact with, but that could also be an application that you interact with it's essentially just like all of the features that are being coded and that being generated by humans right now will eventually be generated indeed by this middleware layer and you will
interact with it and you will not even know that you're interacting with it then the last prediction
is that i believe that within the next 12 years, literally every single human job will be replaced.
And all that we will do is become operators of these agents within the next 12 years.
or whatever it might be, our relationship with AI will purely be based on the exertion of command and control.
And that will depict indefinitely whether we are going to survive as the human race or not and somebody mentioned it earlier um there's going to be a large scale event sooner or
later i hope it's not going to be um a negative event like through war or something like that but
we will have to fundamentally rethink our relationship with these essentially um hyper essentially hyper intelligent beings that we're slowly but surely creating
and we will have to
make sure that
my prediction is going to be a
positive one and not a negative one and that we're not
going to end up in a matrix like
scene where we're all in pods
and we're just being harvested
you know
that's the doomsday scenario
obviously everybody on the space
right now i see that we've had like a very good run uh many people are tuned in thank you very
much everybody for speaking i'm sorry that we don't have enough time to let people on the space
and actually ask questions normally we do that but it's already been 20 minutes over time. So we're
going to end it here and I will see everybody in two weeks where me and Marco are going
to host a seventh edition with new speakers and new topics. And hopefully you've learned
something today. If not, then also come back for the next one because we promise that we're
not just going to repeat the same whole thing over and over again so see you guys appreciate it thank you very much
bye thank you bye bye thanks for hosting thanks guys appreciate it
