Projects
![]()
No projects found
Investors
New
DAO Treasuries
DATs
Lend / Borrow APRs
Chains
Categories
Spaces
AI Agents, Billion-Dollar Deals, And The Battle For The Future
Recorded: Sept. 2, 2025 Duration: 1:14:31Space Recording
Full Transcription
Thank you. Thank you. Thank you. Noah, are you there?
Can you hear me?
Okay, we'll wait a bit
Excellent.
Okay.
Hey, let's co-host Lewis, and then let's also, we need to promote the space, Lewis,
because I think people got tripped up by the crash, and get the new link out there.
We'll get started in about, we're going to start gonna start about i'd say five to ten minutes late
that's fine let's just okay blast this link out and then we'll go from there we'll hang out good Thank you. Thank you. Thank you. Thank you. Thank you. All right.
Last, we can cut the music.
I think we can get started.
People are going to start seeing the new link and join up.
I don't know.
Do we have David up on stage?
I don't see him up on stage.
No, don't see him.
All right.
Well, hopefully David comes up.
A lot of stuff to talk about today.
Yep.
I'm trying to get Ryan up here as well.
How are you doing, Lewis?
Good, good, good.
It's been a really busy week, and I think we've seen some pretty big changes even in the last seven days.
Right.
Our agent stuff, a lot happening in that space.
It's a lot of billion dollar deals.
It just seems like everything is a billion dollar deal at the moment.
Seems like there's a real battle for the future around infrastructure and, you know, just getting out there and becoming productive, not just hype anymore.
I mean, we saw there was an article yesterday of Mark Benioff of Salesforce saying he cut 4,000 support roles.
They went from like 9,000 down to five.
We're seeing a lot of multi-billion dollar investments.
We saw sort of the next shoe drop with India on the Stargate project.
And then there's a fair amount of clashing around, you know around consumer safety, direction, enterprise stuff.
So, yeah, pretty busy week.
No kidding.
So what do you think it is about cutting the supporting roles?
You think agents are just doing those jobs?
Yeah, I think we're seeing a lot of customer service getting realigned.
To be honest, I don't think you can eliminate it completely.
I think there still has to be a human in the loop.
But instead of having thousands of customer service reps,
maybe you only need a few thousand with a few thousand AI agents,
and they're monitoring.
In the article that was in, that
Benioff talked about, he mentions that, you know, there are humans in the loop. They are working
with the agents. They monitor the stuff. So it seems like they talked about an omni-channel
supervisor, a real-time monitoring tool, which allows the supervisors and managers to oversee
the collaboration between the
human employees and the ai agents so i i think what we're seeing is sort of that restructuring
where where they're going okay we you know we can bring in the ai agents to do some of it but you
still really need human beings which thank goodness you still need human beings. You know, so I think AI agents seem to be the real hot thing at the moment.
It's like, you know, we had LLMs and models and AI models and all that.
But it seems like the AI agents are really the killer app for artificial intelligence.
And they're really not just assistants anymore.
They're really being operators.
And you're starting to see some patterns emerging around the usage.
And maybe we're seeing the downsizing in the customer service is because that's where there's a lot of ROI that people are kind of applying.
And customer workflows, a lot of stuff with AI agents and workflows this week.
customer workflows, a lot of stuff with AI agents and workflows this week.
And I think, you know, overall executive leadership's kind of reframing the roles
in the customer service area around agents.
And I think that's why we're seeing that being really significant.
The question will be, you know, where do they go next?
I haven't seen Amazon's numbers, but one thing I've noticed with Amazon
is when I'm trying to return an item, in the past, I would have to, if I couldn't get through the automated
answers, I would have to then get connected to a customer service rep and explain what my problem
was with the item and then have it returned. And now what's happening is I'm actually just speaking to an AI
agent, it seems. And I so far have a 100% success rate because I order a lot of stuff on Amazon.
And oftentimes, I don't know what's going on with it. Quick, quick tangent. I don't know what's
going on with Amazon, but a lot of the products, I feel like the reviews are botted and they're
just not, they're not what they're marketed as.
So I ended up returning a lot of things and I have a hundred percent success rate of just
interacting with this agent and the agent refunds me to my credit card.
So I, I would love to see how many customer service reps, Amazon, for example, has 86 and completely moved over,
kind of moved that workflow over to agents. Yeah, it looks like Amazon's taking a different
slice with the whole thing. They're focused more on downsizing the management level customer
service roles. They've trimmed that. They trimmed that in 2024. And I think they're being really careful and selective of what they actually swap in with
AI agents versus people.
So while Salesforce might be so heavily IT that it might be easier for them to do it,
but where Amazon is interfacing far more with consumers who are not familiar with technology,
they're not familiar with AI, they're being far more careful.
You can see the same sort of strategy with Walmart.
Walmart's doing a lot of very careful step-by-step testing with AI and AI agents because we've
seen examples where customers have gone, oh, we're an AI front company and they've lost
their customers, right?
Consumers just hate it. So I think, you know, once people got burned around that, they're going, okay,
where your target market are consumers, let's be really careful. And so I think with Amazon
downsizing in their customer service, it's primarily in the management roles, but not in the,
in the, in the, you know the service person to person type stuff.
Though obviously, you experience it, they make it more efficient, it can be more easier.
So it looks like they're getting it right, at least for the consumer segment.
The top quote of the week, adversarial AI is coming for your applications.
What do you make of that?
Yeah, there's a help in the security section on the newsletter, you'll see there's a bunch
of stuff where bad actors are essentially repurposing open source AIs to create attack
vectors. You know, we saw last week, AI browsers, which were, you know, being having prompt injections and getting taken over and people actually losing money and a whole bunch of other stuff.
So, you know, that's why I kind of said, you know, while there's a lot of hype around and a lot of activity around a genic AI and AI agents, security around that whole kind of focus really continues to raise its head.
around that whole kind of focus really continues to raise its head.
It's like we're going hell for leather, you know, and implementing a lot of this stuff
and security is getting left behind.
But I think we're slowly catching up.
You're seeing a lot more activity by CISOs, a lot more by the regulatory groups trying
to figure this stuff out.
And I think people are finally realizing it's a huge vulnerability and,
you know, we need to figure out, you know, we need to solve that problem.
But, you know, also in terms of LLMs and AIs, you know,
pretending they're your, you know, your daughter's, you know,
car's broken down and she needs a quick $200, you know, that sort of stuff.
And the voice comes in and grandma says, oh my goodness,
let me help you. And it's not, it's some AI, you know, that sort of stuff. And the voice comes in and grandma says, Oh, my goodness, let me help you. And it's not it's some AI, you know, so there's, there's a lot of issues around
attack vectors with AI. Yeah, David, go ahead. Sure. Yeah, I wrote an article about this a
little while ago. And, you know, I think, you know, these type of AI systems, the only way to counter them is by having the individual have their own AI.
That's actively looking out and protecting them against this kind of stuff.
To me, that's the only solution to these extractive AI systems is to people to use the same tools to protect themselves.
people to use the same tools to protect themselves, right? You know, you're going to have 200 IQ
agents coming at you all the time, right? And how do we solve this in the web context, right? You
needed, you know, a firewall, right? You needed to scan for viruses, you needed, you know, the
software actively looking out for you, right? I mean, part of why I use Gmail is it's so good at filtering all the spam. Right. You know, you get all these these type of protections and I want you know
my AI scanning all of these messages that are coming at me just like I would
email and everything else yeah I think we really need the hardware vendors that
you know the iPhones the androids everyone to start going we need to
build in this security we need to create AI secure.
Like you said, some form of AI firewall or AI protector,
an AI avatar security that basically is monitoring everything in real time.
Otherwise, it's like lambs and sheep and wolves coming in and taking over.
Well, and the challenge is, you know,
some of this is coming from inside of the house, right?
Google gets paid to show you information, right? And they'll show you information from whoever the highest bidder is, right?
And so they've sort of got this conflict where, you know,
the attackers are often hijacking, you know,
the recommendation engines or the advertising ability inside Google or YouTube.
And they seem very slow to take down these, you know, advertisements or even videos that are clearly scams.
But, you know, they're driving a lot of impressions.
They're paying, you know, for that access,
right? And so what I think we have to do is sort of flip the script. And if you own a personal AI,
you know, all of a sudden it's aligned to you. It's not aligned to Google or to Facebook or
some third party, right? That can be easily bribed by an advertiser to get this stuff in front of
you, right? And so for the same reason that we all ended up
with a personal smartphone,
we'll each need a personal smart agent that protects us,
and that isn't owned by some third-party company
or other person.
So big credit to all those in the decentralized AI community
that have been building this out,
whether it's the folks in Morpheus or Venice or Akash, others building sort of all the primitives that we need
in order to really roll this out at scale. But for me, that's the main solution.
Yeah, you broke up a little bit at the end there. But yeah, I mean, I'm not sure if we don't get
supply. I mean, I've got ABG.
I know everyone's probably got antivirus on their PC,
their Mac, whatever, to protect themselves.
But I think it's a huge business opportunity.
I think the development maybe by those virus protectors
to suddenly say, okay, we need to step up.
We need to have an AI LLM that protects people and acts in real time.
But then we come back to, well, you know, our current hardware doesn't really support that really well.
And, you know, people are repurposing, you know, huge amounts of compute into attack vectors.
So for the moment, it seems like security wise, it's going to be a little crazy.
And I just want to also talk about the,
we've heard about the Stargate project,
which is a $500 billion Stargate project.
And essentially it's all about
building massive supercomputing facilities
across multiple continents to support advanced AI models.
And the second shoe dropped this week with India
getting huge investment from OpenAI.
But also there's other investments.
Alphabet landed a $10 billion Metacloud deal.
We're seeing sovereign AI grants.
We're seeing gigawatt scale data centers being planned in India to support the whole OpenAI
stuff.
So there's a tremendous amount of finance, venture capital, and also government support
going into massively building up the infrastructure as fast as possible. Hopefully the scaling
doesn't crack and make the whole security thing even worse, but they're certainly running
really, really fast. Noah, what's, what's your thoughts about, you know,
this sovereign AI and the Stargate project in general?
Hello?
We can hear you now.
You went on mute.
Oh, I think we lost him. He'll be back in a second um one one other pieces around it i think we're seeing that that sovereign money is starting to shape the compute geography like we've seen the
investment with india obviously the u.s is really big we haven't seen that much coming out of europe
which is really interesting though you know there's been some blips on it, but not much. China and the Middle East, we've obviously seen a lot
of focus on chip technologies. Yeah, go ahead, David.
Well, I was going to say on Europe, they sort of shot themselves in the foot with their
whole AI act, right? Basically crippling, yeah. Really sort of limited their ability to, you know,
fund and scale projects.
Yeah, I was gonna say the AI act has really caused
a lot of issues in Europe because of all the red tape
and regulation that it introduced.
And so I don't expect much coming out of there.
Noah, are you back with us yet? Not yet. Well, let's keep on rolling. Can you guys hear me?
So we're talking about Middle East and China. The focus there, a lot of pouring money into NVIDIA alternatives. China is actively supporting a whole new chip competition with NVIDIA.
While there was some great articles actually of the black market or grey market of getting
NVIDIA hardware into China.
But the view of those in China is it's just,
we're just bringing product in.
They don't view it as being a gray market or a black market.
They're just satisfying the desire.
And there's a lot of interesting channels of all the GPUs,
NVIDIA GPUs heading into China that's supporting it.
But even so, the Chinese government as a whole we're seeing lots of push to develop their own internal infrastructure
I think there was a huge hardware deal that also got sparked in the last two
days in the news roundup which is really crazy so the other one was Abu
Dhabi was backing a bunch of funding for diversification for chip suppliers.
So there's lots happening in there.
We saw AMD shedding some more light on their GPU stuff.
Intel, I think, is not really in the game.
They seem to be kind of going slow in the background.
So we'll see what happens with them over the long term.
But there's stuff being planned, I'm sure, on their part.
Yeah, so OpenAI.
I want to talk about their one gigawatt data center in India,
and it really does highlight a shift.
And, you know, AI is now really becoming part of the power infrastructure story.
Obviously, local politics and energy availability.
We're seeing, you know, people's electricity bills in the US going up like crazy.
Some people are pointing to AI with the energy that it does have, though.
Google did come out with a paper.
But we'll see how that actually moves along.
I hear you know it.
Spaces is also driving a few people crazy.
So there's lots around that.
David, you still with us?
Any comment?
Yeah, I was going to say a lot of this boils down to regulation.
For some reason.
Can you guys hear me?
And see if I can solve that problem. Yeah, this is like what a ridiculous. Can you guys hear me and see if i can solve that problem yeah it's just like what a
ridiculous product can you guys hear me okay david and lewis so i'm i'm back up no are you still
around uh i still can't hear you i i am i feel like i feel like people in the audience in here
spaces is insane um what an absolute dog shit product, honestly. Power access is really about, you know,
getting deployment at the moment.
We're seeing a lot of things around that.
We're also seeing some pretty huge power purchase agreements
where people are doing long-term contracts
for data center operators,
and they're buying the electricity, you know,
from not only the standard places, but wind and solar and storage projects.
And they're hopefully tying their compute growth to a new clean generation, which is great.
There's a lot of relationships starting to get formed in the AI space with utilities and regulators, you know, adding their transmission and demand response.
Yes, David, go ahead.
Can you guys hear me now?
David, can you hear me?
Yep, I can hear you now.
Okay, Louis.
Still can't hear you.
Sorry.
I think Louis.
Yeah, no.
One second.
I think it might be Louis' audio.
Okay, can do it.
Exit out. Drop be. Okay. Can do it. Exit out.
Drop down.
Okay.
I'll leave and come back in.
Jeez.
I think after,
you know,
three years,
it's been three years now since we,
we were one of the first movers on spaces and it just feels like the product
is getting worse, not better.
It's ironic.
We're talking about AI and all the great things that come with it.
And I'm surprised the team over at X isn't able to leverage AI to build a better back-end code.
On the back-end, are you familiar?
I'm curious as to why so much time has passed,
and the product seems to be getting worse, not better.
That might be different edge cases around the technology.
I know they had a major refactor when they took over the company
and sort of shifted around.
To go back to Lewis's question, I think the main challenge in Europe is the AI Act.
They've just tied down their companies with so much regulation.
It's almost impossible to get anything done there.
And they've only given basically Mistral a green card to go forward
and build things and release things in Europe. And so, you know, we're seeing a lot of innovation
in the US, China, India, but I don't think you're going to see much coming out of Europe. I mean,
it's just the bar is too high to do anything when you have to fill out a bunch of forms and spend
millions of dollars figuring out
whether your company complies with their byzantine maze of rules so you know i'm pretty bullish on us
at this point uh china's come out with some good open source but you know i don't see anybody
moving towards uh europe to build what they want to build right um as far as i know it's really
mistral is the only thing to come out of there and that's been with a lot of uh you cannot hear
david anymore oh man i don't know what is going on with spaces today this is this is it's ridiculous
i can't hear you david i don't know if the crowd can hear you. Bread Guy, can you... Here, let me see. I see Bread Guy in the...
Never mind. Bread, can you do a thumbs up if you can hear us?
Can people in the audience hear... Okay, so...
Edge Lou can hear me, obviously. He's throwing up a laughing emoji.
David, are you able to hear me?
I can hear you.
Can you hear me now?
Okay, now you can hear me.
Fantastic.
Hopefully, Louis can get some hearts.
Thank you, Sharon.
All right.
Let's try to reset because I have not been able to hear anything that you guys have discussed for the last 10 minutes.
So I'll pass it over to Louis, and let's keep it going.
I guess we can't hear Louis.
David, you were mentioning something about Europe, and you cut off for me.
So if you don't...
Yeah, I was just saying, Lewis had mentioned there's not a lot of innovation going on in Europe, and I think it has to do with all their regulations.
Is it worth doing a restart? It's, you know, it's too bad.
We have a good number of people in here,
and it seems like a lot of folks are interested in I'm going to shoot everyone to follow.
Seems like a lot of folks are interested in the AI conversations.
We do these every week, every Tuesday at 1230 p.m. Eastern Time.
That's 430 p.m. UTC. And we're going to continue to do them.
And I think they're going to be a lot more interesting than the crypto discussions that we've been having, especially particularly over the last, I'd say, year.
I feel like crypto's gotten kind of stale.
We're talking about the same stuff over and over.
And so I'm excited.
I think AI is a lot more exciting at the moment.
And it's a lot more mainstream,
right? Everyone's using it in different industries. I'm actually working out of my
two of my best friend's office. They're a software company. They're founding
it for the last eight years, and even they are figuring out how to integrate AI into what they've built.
So I think it's just something that people have a lot to say on.
Lewis, can you hear me? Or David, can you guys hear me?
Can you hear me?
David's back to a listener again. Let me invite him up to speak.
I think that I'm surprised that X has not leveraged AI, somehow leveraged Grok.
Hey, Grok, make spaces an impeccable user experience.
Wouldn't that be something?
All right, let's bring up.
Can you hear me?
Yes, I can hear you now, Lewis.
Can you hear me?
Yes, I can.
Yes, I can.
My goodness.
I wonder if Blue Sky has spaces. Maybe we should migrate. Yes, I can. My goodness. Okay. I wonder if Blue Sky has spaces.
Maybe we should migrate.
But anyway.
Oh, my goodness.
Yeah.
It is wild.
It is wild.
So, yeah.
You're right.
If the product was three years old, you'd think they would have had it done by now.
But anyway.
Well, what I'm thinking, David, is like, you know, you talk about how you use agents for programming.
You know, you talk about how you use agents for programming, and it's beyond me why the engineers at X must be using agents, but why at this point they haven't been able to make this product that brings so many people together and honestly served us so well during the bear market, why they're not able to make this problem.
Yeah, it is kind of...
It was very paid.
You know, in the first year,
a lot of people criticized spaces
and I stood up for them so much.
I said, you know, it's going to take time
for these things to improve
and there's a big leadership change.
But now it's been three years
and the tools like agents to do things like code are abundant.
I just don't.
Anyway, that's my rant.
Yeah, the first 30 minutes of the show has been.
Yeah.
Well, yeah, maybe, yeah, maybe we need a no-code version of Spaces
that actually works, you know?
I don't know.
That would be nice.
But we were talking about two things.
I know David mentioned Europe, and I totally agree with him, I think,
you know, with all their regulatory stuff over there,
and just the fragmentation of the EU.
I don't – he's right.
I think he's right.
We'll see, you know, progress come out of's right. We'll see progress come out of the US. We'll see progress
come out of China. Now with India, we'll see that. But I think Europe's going to be really slow behind
the ball on this whole thing. Though maybe in other areas, they might be able to come, you know, into some leadership.
Some stuff in healthcare popped up this week, a new AI-powered stethoscope, which apparently can detect heart problems in 15 seconds.
So, you know, if you're a Star Trek fan, you remember the old Tricorder.
I think we're really getting to the point where AI will integrate into a device like that.
And, you know, maybe more than heart problems,
it'll start to detect other stuff.
But I think on the healthcare space,
we're starting to see practical solutions being brought to market,
which is really awesome.
Really awesome.
Captain Levi, go ahead.
And he disappeared.
Well, I just wanted to say, I think the other area,
I just want to say, I think the other area that most of the innovation
will come out of is jurisdiction agnostic, right?
So, you know, the innovation,
the last 10 years in payments
has all come out of crypto, right?
It's been stable coins
and it's lived natively on the internet, right?
Nobody would give a home to crypto from 2014 to 2024.
And so blockchain technology
has become the place
where you host all of this stuff, right?
And so as regulations get thrown up
in different jurisdictions, AI is gonna keep progressing.
It's just gonna take the journey of Web3
where it's gonna be hosted on chain, right?
People are gonna create agents.
You're not gonna even know where the developers are.
What do I care if they're in Ukraine or Singapore or Texas, right? And so
that's, I think, the bigger story here, you know, is agents are taking off in Web3 and in crypto
because that's where they can be built and paid for and rewarded permissionlessly, you know.
For the folks in the Morpheus ecosystem, this is the two-year anniversary of the white paper which is really
amazing to think that two years ago some anons dropped a white paper and now three billion
dollars have gone through the smart contracts right and that was all built on Ethereum and base and Arbitrum, right? So, you know, none of that is waiting for permission.
Wow, that's amazing. Three billion. Yeah, I think you're right. I think, you know,
traditional regulatory, you know, evolution is just so slow, depending, you know, given what
we're seeing with the rate of change and progress in AI, the ability of Web3 to actually cope with that rate of change.
And, you know, the focus on open source, given China is a huge, by its nature, a huge open source supporter.
those three are really where we'll see the most progress and where we'll see the most
adoption because it's just easier than the traditional areas.
I mean, we're seeing there's some stuff coming up on finance, banking and finance.
There's still fraud.
There's still data quality issues.
There's still models showing anti-fraud models, however, are showing results.
The US is trying to do some wealth planning based off AI.
There's a whole bunch of stories around that this week, which is really interesting.
But at the end of the day, clean data really beats any clever prompts. So data quality is still a huge problem and is still just defining
the progress in many ways. And Noah's telling us that you and I, David, are cutting in and
out for him, which is probably part of the whole space's issues, but we will continue
on. David, what's the approach with, you know, you said 3 billion in tokens have gone through
on the Morpheus chain.
Can you expand on that?
How is that actually happening on the chain?
Sure.
Morpheus sort of pioneered this idea of a fair launch where people could stake their
Ethereum towards the project.
And so it all started with Lido, right, which is a popular STETH token, $30 billion worth
of people use to basically get the 3% a year from the Ethereum staking reward, right?
And rather than take that capital directly,
what the Morpheus proposal said was,
hey, everybody hold on to your own ETH.
Just stake it towards the project and direct that 3% of yield
and use that to support the open source.
Use that to support all these new agents
that you want to see built, right?
And it's sort of like,
I'm going to hold the Ethereum anyway.
I might as well stake it towards
a cool open-source project that I want to support.
A lot of people have emulated that model.
The contracts were used by Arweave,
Nounspace, and many others that said,
hey, that's a great model.
We want our community to support us,
but we don't want to be a custodian.
We don't want a treasury.
We don't want centralized risk.
How do we do this all in the smart contracts, right?
And so that became the basis
for how you build out this technology.
But it had sort of amazing knock-on effects.
Like, okay, if I can stake my more tokens now
and give inference, give brain power to my agent,
now it's basically lives forever, right?
I can keep those tokens in that account
on an ongoing basis and always have inference available.
It's sort of like we take for granted in Ethereum
that you put a smart contract out there
and it runs on the Ethereum virtual machine forever.
All right.
Anybody can call that contract,
it's running on chain.
We need the same thing for agents.
And that's sort of what this is big focus on.
Awesome.
Awesome.
Yeah, I think you're right.
I think in terms of finance and Web3, we'll see the most significant amount of change. And now with AI starting to integrate with smart contracts, it's really only going to accelerate. I'm working on a project myself where we're integrating AI with smart contracts and both a utility token and a wrap stablecoin so that's going to be
interesting there was the recent announcement out of the Department of
Commerce actually that they're also focusing on some stablecoin activity
which is really amazing that the US government's getting that deep into it
that fast but I think it it bodes well with the innovation and maybe a lighter
touch when it comes to the regulatory stuff.
A lot of things happening in robotics.
I just want to go into that.
It's really starting to move into mainstream.
Humanoid assistance, surgical bots.
Been a couple of stories on surgical bots that have been doing like gold bladder removal
been doing like gold bladder removal and doing it successfully.
and doing it successfully.
Also for the environment, cleaning drones coming up high in some of the news stories.
So I think we'll see more and more robotics and drones and that sort of stuff coming into
the consumer space pretty soon.
We saw in China, they announced a humanoid robot
for what was it, six and a half grand.
So certainly the price point's low,
whether it can be smart enough to actually do any work,
we'll see.
But once you combine robotics with Internet of Things
and all the sensors we have out there,
AI is gonna be the only way to manage it.
I think with Morpheus, I'm not sure if you guys,
do you see any, because you guys are totally on web
and on chain with the blockchain,
but do you see any physical products at all
coming out of anything you're doing
or is it all totally technology
online based, David?
Well, you know, I think embodied AI, which is really what robotics is becoming, is going
to be powered by all these same tools.
Because at the end of the day, you know, let's say I get an optimist.
I want it to follow my directions. I want to give it details about
the layout of my house. I want to, you know, set the tasks that it's going to do for, you know,
the laundry or mow the lawn or whatever that is, right. That's all personalized, right. And so I,
I'm going to be a lot more comfortable with, you know, robots in the home, robots in the office,
if I've got an individualized way to control all of that.
Imagine the attack vector if you've got a third-party software provider controlling your robot in your house.
The first time it gets hacked, what kind of crazy stuff is going to happen?
Your stuff walks away while you're at work. in your house like the first time it gets hacked, what kind of crazy stuff is going to happen? Right.
Your stuff walks away while you're at work.
Yeah.
People are going to want
personalized control of the robotics in their own house.
Today that's delivered through apps and stuff like that,
but we need to get that on web through
Rails so that you've got sort of a much more robust
security infrastructure, right?
Anybody that's been in crypto appreciates
kind of not your keys, not your coins.
Well, it's gonna become not your keys, not your agent.
Yeah, not your keys, not your robot.
Yeah, 100%.
I think we will have a device which will hold the keys and will allow us to secure the robotic
devices around us.
I think we just have to.
Like you said, people will get murdered by these things, either through action or inaction um and then we'll get back
into the whole three laws of robotics with asimov but um there's also uh been a whole bunch of news
stories around vibe coding vibe coding continues to kind of uh get hyped up um but i think what's
happening is they're now realizing that ship beats I mean, it's time to get real.
And, you know, prompts really can't fix unclear specifications.
I mean, we've probably all found that, you know,
you put in a prompt and what you get back isn't quite what you want.
So you have to recalibrate and refine and refine.
Same thing has to be with vibe coding, you know. And that means you need to have production discipline to get a product out there that actually is going to work.
But that also requires upskilling with engineers.
just announced a whole AI course,
which is about time the universities are starting to try
and get into the whole act around supporting AI
and supporting people understanding AI.
So I know Noah and I talked a couple of weeks ago,
you know, will education in the form we see it today continue?
And I think in many ways it won't.
But what at least we're seeing is that universities and educational institutions are starting
to put in the work and integrate all the AI technology.
I was talking with a trainee teacher yesterday and she said it's interesting because
some schools are offering AI supported education planning as well as AI marking the kids exams.
So it is starting to happen, which is great, but it's certainly going to take a while.
Well, personally, I think that's a huge positive, right? You know, this old 200 year old model of a teacher standing in front of, you know, 25 or 30 students is pretty outloaded at this point, right?
I don't know if you saw the great podcast that the founder of Alpha School did, but they've been switching all of their private school infrastructure to a personalized approach
with an AI tutor, right? We've always known that tutors were extremely effective in education,
but it didn't scale, right? You couldn't get enough teachers to tutor the kid, but now you can,
right? And what they've seen is their students move into the top 98th percentile in just the first year or two in a personalized
AI tutor. And, you know, that's an expensive private school. You know, people pay 40K
to send their kid to Alpha School and their campuses across the U.S. But what they're
talking about is releasing a tool that everybody can use, that homeschoolers and other schools
can access.
And I know some of the developers involved in that.
And to their credit, the university version of that,
Ludwitt is building in the Morpheus community an AI education tutor,
and they've got their first class graduating on Thursday.
So big credit to those guys.
I really love the educational opportunities here because students can accelerate based on how much time they can get from an adult, how much tutoring they can get from somebody that's further along from them. And I think this is a huge unlock.
I love Khan Academy. I think they're awesome. And they've come out with Khan Migo. It's an AI tutor and teaching assistant. And so, you know, with students, it acts like a tutor, you know, guides them through the lesson and gets all that. But for teachers, it helps them to automate tasks like lesson planning and crafting rubrics and all that other stuff that they have to go through. So, you know, it's really helping both sides. And, you know, for those of us who might have had some learning difficulties,
trying to understand certain things, I think it's a godsend. You know, I think finally,
like you said, having a personalized AI tutor to help anyone understand and get over those
initial roadblocks, you know, when you're feeling either embarrassed or stupid, like, why can't I get this, you know, and they go, and they're just
patient, they just keep, just keep working with it until you actually understand it. I just think
it's an awesome time in education. And I think Sal Khan likened AI to having five amazing graduate
students in each classroom. And they support grading, they support lesson planning, they deal with real-time feedback,
and I think they can really act as an extension of teachers and an augmentation of teaching
overall, which is really awesome.
There's some reports that said Khan Academy has partnered with Microsoft to bring it to teachers to the K-12 educators across the US at no cost, which
is really amazing.
They're going to leverage the Azure OpenAI infrastructure to scale it.
But I think that's really just awesome that education is really taking, at least in some
areas, to heart the whole focus on AI and really using it. Certainly research and development's doing the same.
Tons of story in the research category in the newsletter.
Actually, and also some interesting reviews
where AI have gone back and looked at previous
recent reports coming out and how much AI
is integrated into them without anyone saying it really is there,
which is really not good. But at least we can now kind of go back and look at what research
and development is doing and also use it to accelerate research and development.
What else am I seeing? Okay. Yeah, There's some awesome stuff around property. Real estate's
also booming. Real estate and insurance firms are using it for lead qualification and risk
evaluation. We're seeing training programs getting really focused by culture, which is awesome.
But there's still a real need for transparency and traceability. So I think on the on the regular regulatory side, I think
we're seeing logs being a real focus for audits and auditing across the board with logging on
anything that happens with AI. And I think that's probably one of the easiest things to implement as a KPI is logging for AI.
And Web3 should facilitate that.
David, on the Morpheus side, what's your view on auditing and logging with Morpheus?
Obviously, the chain's public.
You can see what's going.
But are there any tools to actually dig in and understand what's going on?
Yeah, I actually think that's an area where agents are going to be really powerful,
is it's not just, you know, coding something up, it's being able to audit the code that gets written.
And if you talk to these auditing firms, most of them are using an AI tool at this point
to scan the code and the library and check for existing and known
vulnerabilities. So, you know, I think that's actually going to be one of the more valuable
things is, okay, there's a million agents out there. How do you know which ones are good?
How do you know which ones are safe? Right. And so having a community like Morpheus, it's just
focused on basically curating all the agents.
You know, you can see on chain their reputation.
I really recommend folks in the Ethereum community to check out the new 8004 ERC that proposes a registry for agents for their identity, for validating what they do, and then keeping their reputation.
And I think those three primitives, those building blocks, are what everybody's going to need in
order to build sort of safe and verifiable agents. I'm not going to want to grab something off the
shelf. We're seeing these exploits where somebody is running an injection script or some other malicious code, right?
And people are just grabbing the tool they saw online
without really knowing if it's safe.
So I think that's going to be one of the big steps here.
Yeah, the 80004 looks really interesting.
They're calling it trustless agents,
which is a really interesting approach
where you've got the agent's unique agent ID is stored and it's used to calibrate and I guess verify identity.
So you have agent registration, which is also then rolled up into task execution and validation.
That's really awesome.
Yeah, this looks like an
amazing thing. And I think this could go a long way in terms of solving some of the security
issues we're seeing that are popping up across AI agents, you know, everywhere, as it seems to be
getting used more and more and more, but people aren't really kind of paying attention to the
security side. Another interesting story that popped up this week was the first music creator
who has no background in music and no musical talent actually signed a record label.
They signed a record deal, which is kind of crazy.
And it goes back to what Nora and I were talking, I think, last week,
where these tools are giving people without the necessary I guess base
abilities like you're not a musician but you can still create music and so in the
creative areas I think we'll start to see a lot more of that but you know this
the questions come up around rights management you know and you know do we
need to watermark stuff and you know we obviously have to say, yes, this is in partnership with AI, I guess you
might say, or in concert.
But there's a real need for consent and, you know, credit and compensation to get figured
out.
But it is interesting that we're now starting to see, at least in music, non-musical people
signing record deals, which is pretty crazy.
The other story that popped up as well, a lot of investment coming up in wearables,
Meta, Ray-Bans, the glasses, AI glasses just raised over a million in pre-orders in three
days, which is crazy.
So there's a lot happening in that space and i
think we'll end up you know eyes up and screen down um maybe we'll even see over the next five
years a full transition you know from our phones to glasses themselves certainly apple seems to be
working diligently on this space um and i think as the brand leader in terms of making things simple, easy to use and no complexity,
they'll definitely be leaders in that space.
But there have been a number of reports of significant investments and raises in the space,
as well as the UCLA engineers have developed an AI powerable powered wearable brain computer interface and so now they can
actually translate your inner speaking thoughts to external which is kind of
wild so we'll see I think with the combination between AI glasses and
we'll also see a greater adoption of AI avatars
because I think in many ways, in an interesting way,
they do go hand in hand.
But we'll see on that.
Again, I think Web3 is the way to go
to keep these things safe.
David, have you heard anything around avatars or music
or anything like this in your space?
Sure. You know, I mean, basically we're lowering the barrier for anybody to become a musician.
Right.
Just like the last year, you know, using something like like GenSpark, which is a wrapper of Claude.
He was able to not only build a website, but take an AI key from Orpheus and hook it in.
And all of a sudden, he's got an AI tutor that he built for all his friends.
And that literally took about an hour of effort
and it took him through the whole process, right?
Creating the website and also hosting the website,
deploying the domain name, every step involved,
including connecting the AI to it, right?
So it's gone from, you know, two years ago
where this was a toy, right?
And it was kind of like nice little demo
to a teenager being able to build and deploy a website in an afternoon right and that's that's
really amazing right because i remember in the late 90s or early 2000s people would pay hundreds
of thousands of dollars for a website they had to host their own servers it was you know three
months of development and you needed all these coders.
And to see that barrier come down, we're going to see an explosion of creativity from all
the musicians who didn't have the technical ability but had something they wanted to say
or something they wanted to create.
As those barriers come down, you're just going to unleash billions of minds to do all these creative works, which to me is really exciting.
Yeah, it's also, go ahead.
You guys have hit the nail on the head, like 100%.
I have an interesting example of that.
I just put this video in the Jumbotron, if you guys can take a quick peek at it.
I just put this video in the Jumbotron if you guys can take a quick peek at it.
This is something that, so by trade, you know, I'm a, I hate to say it this way, but I'm an Ivy League neurosurgeon, so I'm like, I've got a bit of credibility.
I don't have a background in coding, but there's a serious problem in medicine where we can't show people their imaging easily.
but there's a serious problem in medicine where we can't show people their imaging easily.
So with the AIs, I built this, a fully hand-tracked Unreal Engine 5,
most advanced medical imaging software in the world right now.
And I did it all without a great background.
I just listened to the AIs in detail and built it all from the ground up.
So just wanted to share that as an example.
Wow.
Looks awesome.
Yeah, that's very, very cool.
Have you been able to share it with your peers?
I mean, what's the feedback?
Yeah, no, I mean, I formed a company around it.
I have a board.
I have 11 letters of interest from Harvard, Columbia, Cornell,
Sims Murphy, from everywhere, Cedars-Sinai and LA, everywhere.
So I think there's a great deal of desire.
I think I've hit product market fit,
and I think I've hit software design here for an excellent demo.
I guess I'm just at a point now where I need capital to hire
and expand it and build it in a commercial way.
But I just wanted to show it to you guys as an example of like,
hey, yeah, you're unlocking people who were locked up before there's no way i mean to build this would take like
knowledge of assembly c unreal engine i mean like so many different senior engineers would have to
get together to build this uh but i can build this all with uh the ais which is crazy. Yeah. Yeah. I mean, it's like you said, multi-platform access and competence that we've never had before.
It's just astounding.
And so, yeah, so that's wild.
That imagery looks really cool too.
Well, awesome.
Awesome.
Some other things that have popped up in the news,
we're seeing, well, yeah, synthetic identities, who owns your up in the news. We're seeing, well,
yeah, synthetic identities, who owns your doppelganger. We're seeing a lot of things
around that. We're also seeing in the marketing space where deepfake influencers are getting
out there pretty wild and it's becoming a liability for many people. So that's fine.
Go ahead, Giller.
Your hands still up? No, no. Oh, sorry. I can put that down. But I mean,
here, let me see if I can put that down for you. There you go. But yeah, this is something that took me, I came up with the idea on April 29th. It was probably from August 29th. So
in that timeframe, I built something that would normally take two and a half million dollars by myself while being a full time neurosurgeon.
So, I mean, this is this is our future now, you know.
And what what exactly does it do?
I mean, explain it for the most of us who are non-medical.
Yeah. If you get an image like a CT or an MRI and the surgeon is going to cut you open, don't you want them to explore it without a headset first before they do surgery?
Don't you want to know what they're showing you before you get cut open?
I mean, that's what I thought was the issue there.
So they couldn't do it before this?
I mean, we could show you a black and white image or like a 3D reconstruction on a monitor, but this is a 3D monitor with your hands.
So you can see it like a hologram and move it around yourself.
Okay.
And you can pull the slices apart, like from the MRI?
Yeah.
Yeah, yeah.
I mean, you can see it in the video.
Oh, cool. Well, and so does this solve the problem of their allowing them to actually better understand what they're doing before they're operating so that when they operate, they're more either efficient, accurate, or, you know, it works better? How does this actually add to what they're doing? So I'm a surgeon. I've been using virtual reality to plan my surgeries for six years now.
Oh, cool.
Makes me faster, much safer, and honestly just way less stressful, much more enjoyable surgery because I already know where everything is, right? There's no searching, right?
And I was really annoyed by wearing the headset, and I can't get anybody else to use it because
they hate headsets. So once I got a 3D monitor that was holographic, I was like, oh, let me just put in virtual reality controls in here.
And ta-da.
And what's the brand manufacturer for the 3D monitor?
Samsung.
Oh, okay.
Okay.
So consumer level cost, maybe?
How much?
Yeah, yeah, completely.
Yeah, entirely consumer level.
I mean, I built this entirely with consumer level equipment and integrated it.
It's just the code base that's unique, I guess, because it's all built zero day or de novo in Unreal Engine with hand tracking.
Awesome. Wow.
So, yeah, yet another example, you know, people who are highly skilled, but not multi-domain and the ability to go cross-domain and build stuff that we've never built before.
Yeah, I'm sure, you know, this is an awesome example.
I'm sure it won't be the last.
Yeah, I'm a brain and a spine surgeon.
And this last image I just sent, I just sent over another one to the Jumbotron.
That's a chest.
So, I mean, that's how good the chest looks
and the heart looks it's crazy wow wow that is awesome awesome well thanks for sharing that's
really powerful and i think really amazing as we move forward you know with ai and all the changes
that we're getting that's just awesome um though we're also seeing you in different areas causing maybe not so much, more dislocation
and disruption than actually integration.
Courts and law firms are currently testing AI for discovery.
They're using it in contract review.
And now we're being told that even judges are using it in sentencing support.
Not sure how much I like that, but we'll see.
Okay, Noah just sent me a message.
Can I ask everyone to follow the Moby Media account and the speaker accounts, please?
So yeah, everyone please follow us, myself and Noah and Captain Levi, he's speaker, and
Galal who's been giving us this introduction to the amazing
uh 3d stuff um awesome thanks guys and sharon sharon sorry um what else i think i think what
else we got going on in the news uh npcs are finally having memory uh which is awesome and
NPCs are finally having memory, which is awesome.
And, oh, been many, many stories on AI and mental health.
This is an interesting area.
You know, people have been using chatbots as therapists,
and, you know, some have found a lot of good coming out of it,
and some have not.
You know, it seems from what, you know, I'm not a psychiatrist or psychologist,
but it seems from the reading of the media that people who do have significant mental illness
maybe should steer clear of AI therapists.
But, you know, otherwise, a lot of people seem to be getting some positive results out of it.
Anyone of our speakers here today,
have you had any experience using AI for, you know,
just asking questions about yourself or therapy or understanding stuff?
Anyone had any success or maybe issues?
Good morning, everybody.
Hi, Lewis.
I'm so sorry.
David wanted to, he just unmuted so he can go ahead
and I'll just go after. Well, I was just going to say, I'd really want to know the values and
worldview and perspective of an AI counselor, right? Just like I would want to know that
if I was taking, you know, one of my kids to a normal human counselor, like, you know, what are they advising?
Right. And so, you know, I think that's the important thing, sort of just transparency and, you know, the system prompt and how these systems are working.
Right. Absolutely. Absolutely. Sharon?
Yeah, I'm just wondering where this is all going to go. I personally don't use AI. I did, I think, today for the first time, I asked something personal, but I usually don't. I do know that many do.
Yeah, I'm just wondering where this is all going to go. I personally don't don't use AI.
But recently, you know, someone committed suicide and now they're suing OpenAI. And many states now said that you're not allowed to use chatbots, or at least psychologists and so are not allowed
to use chatbots.
So I'm wondering where this is all going to go.
Yeah, I think Illinois came out definitely, and there was a report on 44 other attorney
generals also looking to ban it.
It's interesting. In the distant past, I studied and trained
neuro-linguistic programming,
which is based on three psychiatrist's work,
Virginia Ceture and Fritz Perls,
and someone can remind me of the third one.
So I actually spent some time creating a context
with one of the ChatGPT sessions and giving it instructions around those three and then posing it a problem.
And so give me the answer based off, you know, those three worldviews.
And it was really super interesting, the feedback it gave.
It was very insightful in many ways. Galal, as a medical professional, what's your
view on using AI from a mental health therapeutic point of view?
I think it's important right now that we always have a human in the loop for now. I think people who... The issue is the AIs are not to be trusted as personalities,
and they're not to be trusted alone.
In augmenting a professional, they're actually extremely powerful
because what they do is they help that professional weed through the haystack
and get to the needle more quickly, so the throughput is better.
That type of supervised cobot relationship
will eventually build a data set and an annotated data set necessary to have AIs that we can trust
without humans being in the loop. But we're not at that point yet.
Yeah, I totally agree. You know, I've seen so many stories of people saying, you know,
when they've gone to Chachipatine said, you know, when they've gone to ChatGPT
and said, you know, prompt, create me this, do this, and it's just been crap output.
But when they actually create a draft form and some structure, and then they give that
to ChatGPT and say, review this, tell me what you understand from this, and suggest, you
know, where there might be any gaps, then it actually produces really
amazing and powerful output.
So as a co-support, as an intern type of support, I think it's just amazing and it does really
become a force multiplier for what you're doing.
But I think if you, and I've always said this,
if you don't have the domain knowledge, skills,
and experience surrounding the prompt you're asking,
then you're probably going to get garbage because you won't know where it's trying to BS you
and where it's falling off the rails
and where it might be hallucinating.
So I think that's the real key thing at this point in time,
at least in the overall evolution of what we're seeing with AI models and with LLMs.
And so it's now 40 minutes past the hour.
I'm actually going to probably close out a little bit early just because of some of the technical issues we have.
But let me just wrap up.
For this week, for the last seven days, I think I've really seen AI moving from hype to habit. We've seen agent workflows really proving that they're a real killer feature. And it's really about the agent and not the model to some extent.
out there robots becoming opex you know where automation is now a subscription like utility
manufacturing is certainly starting to do that on a big deal and also starting to set up power
contracts uh for long-term support for electricity and and so on and these systems are really getting
embedded into our everyday life very very fast um but i think the challenge that we we face is that scale really breaks moderation
and and at this point consent needs to become a common core of integration and as we build with
transparency and trust i think then innovation can really uplift people not just the platforms
we're on um i want to thank all the speakers today. Thank you very much.
Apologies for our technical difficulties.
Hopefully we'll get those worked out next time.
And we'll see you again next Tuesday,
clear and free and hopefully much more technical support.
So thanks everyone.
Thanks, David, Noah, Galau, Sharon.
Thank you, everybody.
I'm sorry.
I could barely hear anything.
A lot of technical difficulties, but I'm thankful for Louis and David and the rest of everyone for carrying the space.
Please make sure to follow all the speakers.
Make sure to subscribe to Louis's newsletter.
He puts out some great stuff there.
And also, please make sure to follow the MobiMedia account.
We're going to be doing this every week and hopefully
get bigger and better speakers on to join the conversation and talk all things AI. I think
this is one of the most exciting things that's happening in our lifetime. And so I'm eager to
have discourse and I'm eager to learn and eager to integrate it into my daily life.
Take care, everyone. Thanks, everyone.
All right, guys.
Thanks, everybody.
Keep being awesome.
Bye. Thank you.
