Exploring the Value of Knowledge in Web3 | #Consensus2024 AI Summit Recap
Summary
TLDRThe video script discusses the challenges of data privacy and monetization in the AI era, emphasizing the importance of decentralized AI. Shashank Shara, co-founder of GYET, explains how decentralized AI allows individuals to own their data and contribute to learning models without reliance on large corporations. He advocates for a system where people can monetize their intellectual property through AI agents, facilitated by web 3 technology, to prevent monopolies and empower creators. The conversation highlights the need for transparency, attribution, and fair compensation in AI development.
Takeaways
- 😐 The current challenge with AI is the issue of data privacy and monetization, where individuals give away their data for free, which is then used and monetized by others.
- 🤔 Decentralized AI is presented as a solution to data ownership, allowing individuals to control their data and get paid for its use, contrasting with centralized AI where large corporations control the data.
- 👥 The concept of AI being more than just a tool and forming a relationship with humans, akin to a parent-child dynamic, is discussed as a way to build trust and interaction.
- 💡 The importance of transparency in AI data sources and usage is highlighted as a way to make people comfortable with AI and to encourage its adoption.
- 💰 The discussion suggests that the incentive for decentralization in AI comes from the potential for financial gain, similar to how Bitcoin gained momentum.
- 🌐 Web 3 is identified as a crucial component in the decentralization of AI, providing a framework for attribution, validation, and monetization of data.
- 📚 The script touches on the devaluation of intellectual property in the internet era and suggests that decentralized AI could restore value to knowledge and intellectual property.
- 🚀 The potential for a new class of knowledge workers who can monetize their data and intellectual property through AI agents is proposed.
- 🛡️ The script warns against the rise of new monopolies in the AI space and emphasizes the need for open-source and web 3 technologies to prevent this.
- 🌟 The discussion concludes with the idea that an 'arms race' in AI should be about outperforming centralized models, with the goal of creating a more equitable and decentralized AI ecosystem.
Q & A
What is the main problem with AI and data monetization discussed in the script?
-The script discusses the issue of individuals giving away their data for free, which is then monetized by others without their consent or benefit. This was a significant problem in the web 2 era and is expected to intensify in the AI-driven world.
What is the definition of decentralized AI as per the script?
-Decentralized AI is defined as the ability for individuals to own their own data. It involves everyone contributing to a large learning model, rather than a single giant corporation controlling and training the model in secret.
How does the script relate the concept of decentralization in AI to Bitcoin and Ethereum?
-The script draws a parallel between the decentralization in cryptocurrency networks like Bitcoin and Ethereum, where nodes are run by various participants, and the concept of decentralized AI, where data contribution and model training are distributed among individuals.
What is the vision of AI presented in the script, and how does it differ from current AI models like GPT?
-The script presents a vision of AI that fosters a deep relationship with humans, where AI aids and enhances human capabilities rather than replacing them. This contrasts with current models like GPT, which are sometimes seen as monolithic and less personalized.
How does the script suggest incentivizing decentralization in AI?
-The script suggests that incentivizing decentralization in AI can be achieved by aligning human good with human greed, similar to impact investing. By allowing individuals to own their AI agents and nodes, and by monetizing their data, people can be motivated to contribute to decentralized AI systems.
What role does the script suggest for Web 3.0 in the future of AI and intellectual property?
-The script posits that Web 3.0, with its emphasis on blockchain technology and decentralized systems, can play a crucial role in attributing, validating, and monetizing intellectual property in AI, thus restoring the value of knowledge and preventing monopolization.
How does the script address the issue of AI and the potential devaluation of intellectual property?
-The script discusses the devaluation of intellectual property due to the internet and suggests that decentralized AI, combined with Web 3.0 technologies, can help restore value by ensuring proper attribution and monetization of contributions.
What is the script's perspective on the relationship between humans and AI?
-The script envisions a relationship between humans and AI that is akin to a friendship or a parent-child relationship, where AI is individually trained and trusted, rather than a top-down, prescribed relationship.
How does the script view the role of open source in decentralized AI?
-The script views open source as essential for decentralized AI, as it allows for the development of AI agents and nodes that individuals and institutions can own and control, preventing the rise of another monopolist organization.
What is the script's stance on the importance of transparency in AI and data usage?
-The script emphasizes the importance of transparency in AI, stating that for AI to be truly beneficial and trusted by humans, it must be clear where the data is coming from, how it is being used, and how individuals can benefit from it.
What challenges does the script identify in the current AI landscape regarding data and intellectual property?
-The script identifies challenges such as the monopolization of data by large corporations, the devaluation of intellectual property due to the internet, and the lack of transparency and control over how data is used in AI models.
Outlines
🤖 The Challenge of Data Monetization in AI
The first paragraph discusses the complex issue of how individuals freely share their data, which is then monetized by others without their consent. It highlights the difference between web 2.0 and the potential amplified risks in the AI-driven future. The speaker introduces the concept of decentralized AI as a solution, emphasizing the importance of data ownership and the role of AI in building trust and relationships, rather than replacing humans. The conversation also touches on the challenges of incentivizing decentralization and the need for AI to be transparent and beneficial to its users.
🌐 Embracing Decentralization in AI for Intellectual Property Rights
The second paragraph delves into the practical implementation of decentralized AI, using the example of Gyana, a platform that allows individuals to set up AI agents and nodes. It discusses the importance of aligning human good with human greed, ensuring that creators and researchers are rewarded for their contributions to AI models. The paragraph also addresses the devaluation of intellectual property in the internet era and the potential for a new class of knowledge workers who can monetize their data through AI agents, reflecting on the societal and economic implications of these changes.
🛡️ Combating Monopolies with Decentralized AI and Web 3.0
In the third paragraph, the discussion shifts to the role of web 3.0 in attributing and validating contributions to AI models, which is crucial for monetization and preventing copyright violations. The speaker argues that the combination of open-source AI and web 3.0 can create a fair system for intellectual property rights, contrasting it with the monopolistic practices of web 2.0 companies. The paragraph also touches on the importance of human involvement in training AI models and the potential for a more equitable distribution of wealth in the AI industry.
🔄 The Future of Decentralized AI and Its Impact on Society
The final paragraph wraps up the conversation by emphasizing the necessity of a decentralized system to handle the vast amount of work involved in AI development fairly and efficiently. It stresses the importance of preventing monopolies and the role of open-source AI and web 3.0 in creating a more equitable and transparent AI ecosystem. The speaker also reflects on the societal impact of AI, suggesting that a decentralized approach can lead to a more inclusive and democratic use of technology.
Mindmap
Keywords
💡Decentralized AI
💡Monetization
💡Data Privacy
💡Centralized AI
💡Intellectual Property
💡Web 3.0
💡Open Source
💡AI Relationship
💡Monopolies
💡Knowledge Workers
💡Copyright
Highlights
The current challenge of data privacy in AI where individuals give away their data for free while others monetize it.
Decentralized AI as a solution to data privacy issues, allowing individuals to own and control their data.
Decentralized AI defined as a system where everyone contributes to a large learning model rather than a single corporation controlling it.
The importance of transparency in data handling and the potential for individuals to get paid for their data contributions.
The analogy of AI as a life form that builds trust with humans, rather than a top-down relationship.
The role of AI as an aid to humans, not a replacement, and the need for a deep human-AI relationship.
The necessity for AI transparency to ensure user comfort and contribution to AI systems.
The comparison between the arms race in Bitcoin and the incentivization needed for decentralized AI adoption.
The idea of aligning human good with human greed in AI development, similar to impact investing.
The concept of open-source AI agents and nodes allowing individuals to monetize their intellectual property.
The potential societal impact of decentralized AI in creating a new class of knowledge workers.
The role of web 3.0 in restoring copyright and intellectual property rights through blockchain technology.
The discussion on the future of copyright in the context of AI and the internet's impact on devaluing knowledge.
The importance of avoiding monopolies in AI to prevent exploitation and to ensure fair compensation for data contributions.
The potential of decentralized AI to empower individuals to monetize their data and intellectual property.
The need for open-source and web 3.0 to work together to prevent monopolistic control in AI.
The human involvement in training AI models and the potential for fair compensation through decentralized systems.
Transcripts
one of the trickiest most vexing
problems facing AI right now right now
is how we all give away our data for
free and then others monetize it and use
it and do God knows what with it that
was already a massive problem in web 2
and it could be on steroids in the AI
world right it's one thing when we give
away what type of sneakers we buy online
or where we live but imagine the data
getting more sensitive and more private
it's going to get trickier and trickier
so to dive deep in this issue and to see
how decentralized AI can be a solution
we have no one bowly the breakup and
Shashank Shara co-founder of gyet thanks
gentlemen enjoy thanks everyone thanks
for being here have a seat Shashank have
a seat all right great so indeed you
know the the spirit of the whole event
of course is decentralization now we
understand what decentralization means
in terms of Bitcoin and ethereum and you
know nodes and who can run them and
who's propagating the network we all get
that but decentralized AI is kind of
different what what is it what is
decentralized AI versus centralized AI
yeah I think uh so decentralized AI is a
lot of buzzwords obviously um we we're
defining it as being able to own your
own data uh Ai and large learning models
uh are trained on all kinds of data
they're contextualized and inferenced
all kinds of ways uh but decentralized
AI uh we think is kind being able to own
your data everyone uh acts as a village
to contribute to that large learning
model it's no longer some giant
Corporation that's trained secretly and
now they're ready to go I don't want to
name any names but you all know who it
is and for us it's being able to take
charge of that data be able to uh I
figure out where it's coming from
attribute it validate it and then get
paid for it hopefully now when I think
about why it matters I think about the
relationship we have with AI in a sense
because it's kind of like it's
artificial intelligence so we build
relationships with it you know Bitcoin
is about securing commercial
relationships and being able to transact
in a way that we can have more trust
right because we know the network is
decentralized now you know I wrote a
book on AI and Bitcoin and it was all
about this idea of us having an eyeball
an a eyeball and indeed in my story
they're each individually trained and
we're able to trust where the
information is coming from and we're
able to make these kind of friendships
with them almost like a or even a parent
child relationship does that Vision make
sense to you was was my story on its
right track instead of instead of some
Professor top down relationship that's
being prescribed to you you're able to
raise this sort of life form and
of course I mean I mean think about AI
right I mean our our recent relationship
with AI has been through chat GPT right
and that was released to you maybe you
guys interacted with it starting about a
year and a half ago uh there's others
who maybe are in the test net now when
it was released it was presented as uh
Skynet right it was like Hey we're ready
to go but that's not what AI is supposed
to be or at least we don't believe that
the future of AI we think that AI needs
to be uh have a deep relationship with
humans it is here to help humans not to
replace humans it's to Aid not to uh
replace so how do you do that well you
have to go back to kind of how humans
even interact right humans are uh social
people uh as you said a family bond uh
you're the way to make people feel
comfortable uh with AI and then
contribute to Ai and utilize it for
their own lives is to have uh make sure
that it's transparent on where that data
is coming from how that data is working
and then have a stake in actually
getting paid from it using it and kind
of benefiting from it well now that's
all well and good but we're in an arms
race here right in the same way that
Bitcoin was able to gain momentum in the
arms race it was able to do so because
people made money it wasn't because we
nagged a bunch of people to love
decentralization that it worked out
right what we were able to do was make
more money than the people who didn't
get decentralization and then eventually
they're like oh oh we should do it that
way now how do we win the arms race how
do we incentivize decentralize because
right now I use chat TBT it empowers my
own writing it empowers my own
communication but there's things about
it I don't like I have to trick it
constantly I have to hypnotize it to do
what I want I've got to motivate it I've
got to you know there's all these crazy
things that I have to do to get it to do
what I want so how do we do this with
training these models how do we get it
so individuals want to train the models
and and really cultivate this
relationship yeah I mean I love for
example I I love uh impact investing
right I love a and the way we I always
approach that is aligning human good uh
with uh human greed with doing good uh I
think that's the same for uh AI as well
uh so over Guyana what we're doing is uh
allowing folk to open source set up
their own uh AI agents their own nodes
and we were our firm was born out of uh
project at Berkeley our professor our
CTO Dr Allen Yang he he started to build
virtual Tas at Berkeley and the
reasoning behind that was you know he
wanted he had a hard time hiring people
to be honest for TAS but he couldn't go
to
GPT because they would take his data
they it's expensive it's slow uh when so
he started building his own node and
then permissioning access and for a
researcher that's a big step you cannot
the moment year he he put 20 years of
his research on a node that's his life's
work and so obviously he should he can
and should get paid for it um and really
I think it it pointed towards uh uh a
kind of a wrong that even the internet
perpetrated on intellectual property uh
the internet was has been a really great
don't get me wrong it proliferated
knowledge but at the same time it
devalued that knowledge right and when
you devalue the knowledge uh that's when
you start to see a lot of societal
problems you see newspapers shift to
more uh clickbait uh uh material you
start to see uh ad you know kind of a
more ad-based Revenue just a survive and
so many newspapers went bankrupt and now
you see the the Pinnacle of
it uh these news newspapers and IP
holders are starting to sue open AI now
because open AI has and others have
trained on a hundred years of their
newspaper workor and other IP to get to
where they're at but it's not a specific
model it doesn't really like you said
you have to train along uh we believe
fine-tune learning models uh fed by uh
nodes that are
uh individuals of uh companies
institutions who have worked hard to
create their knowledge uh is the way
forward so in our Ideal World the last
generation was you know YouTube allowed
creators to proliferate and have a
middle class set of creators unlike
before in Hollywood where it was the
halves and the Have Nots you have a
thousands of YouTubers making good money
we think the next stage of allowing
people to own their data and monetize it
will create another class of well-
earning knowledge workers it could be an
individual just blogging at home writing
at home writing poetry writing songs
whatever that that IP is they can
monetize it and that monetization looks
like in the form of an AI agent that
contains their knowledge and they can
permission who gets to access it and for
how much well you know I'll be the
devil's advocate for a second because my
own ideas have to change when I when I
graduated from law school my thesis I
wrote it on the sort of future of
copyright a lot of issues there was Gene
patenting at the time that was looking
terrible right that that that fell apart
thankfully the Supreme Court struck that
down um but my prediction in the paper
which did well I got an A probably the
only a I got in law school um the theory
was that the the copyright came in with
the printing press and was about to go
out with the printing press now this was
before there was AI it was before there
was Bitcoin um now I still don't know
what the future of copyright is uh but I
think things have changed can it go back
can can copyright become an institution
again like it was because it's melting
away right I mean you look at music
sampling no one it's not offensive now
to copy you know 30 years ago a cover
song was like a cover band it was like
yeah I want the real thing right
nowadays everyone's so used to it it
doesn't matter right that that whole
thing became wishy-washy you know how do
we
or do we find new ways to to make money
off of of intellectual property I'm not
sure what the answer is how do how do
you fit into this and how ises AI fit
into yeah so I think that's where kind
of web 3 actually fits into it uh AI I
mean we look on our side we're just
developing really great toolkits to help
you own your data the web 3 component is
absolutely crucial to restoring the
copyright use mention because the
problem is right now once again with web
2 and what's happening is you don't know
where
different components were used in the
case of a music sample you know they mix
too many bits up to properly track right
so that's where web 3 is crucial to the
story you need attribution you need
validation to lead to monetization right
so if you can attribute if everything's
on chain and you know who contributed
what if New York Times is starting to
permission a specific newspaper headline
for to to feed a Al algorithm or a
learning model then it they'll know how
many times it was uh inferenced and uh
what they should get paid uh you put
everything on chain you can't mess it up
and basically if New York Times thinks
there was a VI violation then you can
use our normal court system you can go
to through the normal Court copyright
courts patent law and uh and sue them
and you'll know exactly how many
violations were so if you can at least
track it I think that's a solution to it
yeah perhaps because my my theory was
the old Canadian media slogan the media
is the message and when the internet
came the message was the physical
isolation or the physical um you know
one of one or whatever the that it
wasn't in in bits and bites
automatically meant that the medium
paper books printing had value of course
we took that away now the way you
described it goes back to what I said
earlier this is an arms race yeah right
just like in Bitcoin there's an arms
race here and in Bitcoin I see the arms
racist trying to get richer than the
Fiat people so that people will say well
those are the ones I want to follow it's
not an intellectual thing right yeah so
is that is that what's going on are we
do we find a way to make all these AI
people stand out and if you use
copyright this way you're just going to
be able to lead the way because you're
successful yeah I mean it should be an
arms race uh but the Army the armies are
you know centralization versus
decentralization I mean for example
we're an open source decentralized
project you guys can go download us and
we're not trying to take money from that
um but the Army here is is you guys it's
it's people who actually exist in in
real life uh it's not just some
ephemeral uh training bot uh so yeah you
need to arm up and I think you know the
there's many open source llms now uh
that fed by uh your nodes hopefully
whether use us or set up your own uh AI
agents I think that becomes the Army but
uh the key here is we cannot allow
another monopolist organization to be
between us and the end product uh we
like it we it love it yeah we love the
Applause thank you thank you thank you
uh the key here is the web to has seen
the death of so many companies so many
small and midsize potential businesses
because of the monopolist and I it's
great look we I use every Fang stock
ever i' I've I'm invested in Facebook
Netflix Google Amazon like don't get me
wrong it's made life really easy but
when you're talking about owning your
data and then getting paid for it uh and
contributing to it you we need to remove
the the the centralized player you
cannot do it in web 2 because it'll lead
to the rise of another monopolist and as
you see in Amazon and the others they
will wait till their time or just like
uber they will wait till their time and
charge you three times five times more
and and you'll have to pay it at that
point they've driven the taxi company
out of business you have nowhere else to
go and the last thing I'll say is that
uh you need open source and you need uh
web 3 to work together to combat that
yeah because you know something I
learned at this show in another AI talk
and and I didn't know this that that the
big breakthrough that chat GPT had was
they figured out that human sorder
were a big part of the language training
models they pay people in the
Philippines to say dog cat dog cat until
for hours and hours and hours yeah um
and there you know it's a job like you
know it's not bad uh and I heard that
chat GPT spent you know they're going to
spend a billion dollars this year on
people just humans
going so indeed you know we need humans
to put this data to to tag it to to
upload it into a node and and hopefully
make more than the $1 or2 hour the folks
in the Philippines are making for doing
this stuff yeah you know that would be
great yeah I mean I think that's where I
think web 3's already solved that issue
by by staking by uh you know validating
someone's work um that's like at the
core of web 3 so uh yeah I mean in
funnily enough I mean it's funny we uh
it's it's two big buzzword you know we a
lot of people interested AI web 3 how
does it fit together are we just driving
up a bunch of money uh and a bunch of
people hopping on it it it's but when
you actually go under the hood of uh how
humans uh interact uh to form Ai and
form learning models and neural networks
uh it's crucial to uh have a
decentralized system from the Monopoly
uh monopolist angle but also just pure
there can't be a centralized system
can't handle that kind of work uh and
not in a fair way so this centralized
system of vent production can handle the
work of going overboard please please a
warm Round of Applause for Shashank
everyone thank you thank you
Ver Más Videos Relacionados
Episode 1- Efficient LLM training with Unsloth.ai Co-Founder
How Federated Learning works? Clearly Explained|
AI Radioroom Episode 1 | Exploring 0G Labs’ Solutions to Challenges in AI
Verifiable Compute for AI, ML and More: Web3's New Frontier | Alison Haire at SmartCon 2023
"I tested 400 AI Agents, these are the best" - Adam Silverman
Exploring GaiaNet: The Future of Decentralized AI | The Perfect Blend of Web 3.0 and AI
5.0 / 5 (0 votes)