►
From YouTube: Tom Trowbridge: What Decentralized Compute-over-Data Brings to Filecoin | FIL Network Base Austin 23
Description
After providing storage capacity and data-onboarding, the third phase of the Filecoin roadmap is bringing compute capability to the data stored on the network. Tom Trowbridge of Fluence Labs gives us a glimpse into why this innovation will be transformative
A
So
coming
up
next,
we
have
Tom
Trowbridge
of
fluence
labs,
who
is
going
to
give
a
talk
a
bit
about
what
fluence
is
doing
and
what's
super
interesting
about
fluence
is
that
it's
really
kind
of
they're
trying
to
tap
into
the
decentralized
Computing
use
case
of
the
file
coins?
A
This
is
kind
of
like
the
last
sort
of
you
know
the
final
stage
in
the
file
coin
roadmap,
if
you've
seen
kind
of
Wands
kind
of
three-stage
file,
coin
master
plan,
which
is
the
first
stage,
is
building
world's
largest
storage
Network,
which
we've
done
or
decentralized
storage
Network,
which
we've
done
step.
Two
is
onboarding
the
world's
data
and
we're
in
the
process
of
doing
that.
A
We've
got
about
700
petabytes
worth
of
data
on
the
file
coin,
Network
now
and
then
the
final
stage
is,
is
bringing
computational
abilities
compute
over
data
functionality
to
that
to
the
to
that
data.
So
so
that's
one
of
the
things
that
Tom's
been
working
on
I
hope
I
described
that
accurately.
So
so
that's
that
that's
where
you
know
when
we
say
that
storage
is
just
the
start.
A
You
know
the
the
fvm.
The
launch
of
the
fvm
is
so
exciting
because
it
enables
kind
of
precisely
some
of
the
things
that
Tom's
gonna
be
talking
about
here,
really
bringing
it's
not
just
storing
the
data,
but
it's
being
it's
bringing
computational
abilities
to
that
data.
A
A
Did
you
have
yet
slides
right,
okay,
yeah,
so
I
think
they
should
be
queued
up.
You
can
go
ahead
and
use
the
mic
on
there.
B
Yeah:
okay,
thank
you.
That
was
a
terrific
introduction.
It's
probably
one
of
the
most
accurate
and
thorough
introductions.
I've
had
so
thank
you
for
knowing
about
fluency
what
we
do.
This
is
the
ecosystem
that
would
do
that.
So
fluence
is
a
decentralized
serverless,
compute
protocol
and
there's
a
terrific
integration
and
relationship
with
protocol
labs
and
with
filecoin
protocol
Labs
has
been
a
long
time
backer
of
us,
an
investor
of
ours
and
we're
close
with
the
filecoin
foundation
and
the
entire
ecosystem
here.
B
So
we're
happy
to
be
here
and
look
forward
to
a
lot
more
collaboration
going
forward.
So
first
I
want
to
talk
about
what
is
fluence
and
we've
had
to
back
up
here,
so
decentralized,
serverless,
compute,
Network
or
platform,
and
first
we've
got
to
get
some
definitions
right.
What
does
decentralized
mean
and
decentralized
in
different
contexts
and
from
different
who's
ever
talking
about?
It
will
come
up
with
slightly
different
attributes
to
what
constitutes
decentralized.
So
we
think
there
are
three
attributes
which
are
necessary
to
be
truly
decentral
realized.
B
You
have
to
be
verifiable
and
that
allows
you
to
be
trustless.
You
also
have
to
be
unstoppable.
I.E
censorship,
resistant
and
need
to
be
fault,
tolerant
and
to
be
clear,
you
can
be
fault,
tolerant,
but
still
be
not
be
be
sensible.
So
and
you
can
also
be
sensible
and
also
be
fault
tolerant.
So
you
need
to
have
all
three
of
those
and
they
are
not
the
same
thing
if
you
think
about
it
more
carefully
and
then,
what's
a
compute
platform,
it's
effectively
a
cloud,
it's
a
serverless!
B
It's
a
cloud
stack
that
allows
the
decentralized
applications
allows
you
to
build
them
from
scratch
easily.
So
that
is
what
fluence
is,
and
some
kind
of
key
components
to
our
definition.
And
so
what's
some
more
detail
behind
that,
because
that's
a
little
bit
high
level-
and
why
are
we
doing
this
and
the
objective
here?
B
What
we're
trying
to
build
is
the
ability
to
run
arbitrary
code
without
limitations
with
the
same
kind
of
Speed
and
Performance
as
the
cloud
all
right
and
that
doesn't
really
exist
in
a
decentralized
way
right
now,
and
that
means
we
want
it
to
run
anywhere
on
the
edge
on
your
desktop
in
the
browser.
No
consensus
required
high
performance
and
to
be
fault,
tolerance,
secure,
asynchronous
and
obviously
inputs
from
ipfs
filecoin
l1s
l2s
Etc.
So
these
are.
B
That
is
what
fluence
is
doing
for
compute,
because
99.9
of
the
web
is
not
suitable
for
on-chain,
computation,
right
and
so
most
of
the
web.
You
need
to
be
able
to
run
any
computation
and
be
permissionless,
and
also
you
got
to
be
cheap
and
fast
and
that
doesn't
work
with
blockchain
compute,
so
we're
trying
to.
Basically,
we
have
built
a
general
purpose,
decentralized
compute.
That
is
the
what
we've
been
doing,
and
why
are
we
going
through
the
headache
to
do
that?
Why
bother
because
it's
cheaper?
B
Filecoin
has
proven
this
from
a
storage
perspective,
we're
working
on
it
from
a
compute
perspective,
Cloud
lock-in,
you
know,
the
dominant
position
allows
pricing
power
when
you
can
change
your
cloud
provider,
your
compute
provider,
with
no
barriers
and
no
cost.
They
have
no
pricing
power,
so
pricing
Falls,
and
that
is
that
is
what
we
have.
That
is
the
key
piece
here
to
reduce,
reduce
price
and
also
driving
Innovation,
because
if
you
can't
lock
people
in
then
you
need
to
have
offer
other
services.
B
B
We're
going
to
talk
about,
but
web3
still
exists
on
Amazon
and
AWS,
and
we've
built
all
these
awesome
tools,
but
the
wrong
regulator
comes
along.
The
wrong
government
pushes
Amazon
the
wrong.
You
know,
incident
happens
and
we
are
at
tremendous
risk
of
this
ecosystem
being
de-platformed
right.
So
that
is
why
we
need
decentralized
general
purpose.
Compute.
B
You
also
have
centralized
back-ends
where
people
said
you
know,
we
don't
want
to
be
on
Amazon,
so
instead
they
run
servers
in-house.
That
is,
that
brings
you,
don't
run
deep,
platforming
risk,
but
you
run
a
whole
host
of
other
centralization
risks,
which
is
also
something
we
in
this
ecosystem
are
trying
to
prevent
whoop.
B
So
with
that,
you
know,
the
other
other
point
here
is
that
centralized
compute
prevents
collaboration,
and
this
is
kind
of
the
now.
We've
talked
about
the
negative
things,
but
this
is
the
upside
of
decentralized
compute,
where,
when
you,
you
have
some
people
here,
probably
remember
when
you
could
build
on
Facebook
and
you
could
build
on
Twitter
and
you'd
applications
that
existed
on
top
of
these
ecosystems
and
that
led
to
Innovation
and
with
web
with
a
decentralized
compute
Network
and
with
applications
that
exist
outside
of
ownership.
B
That's
hosted
in
a
decentralized
compute
manner,
and
so
because
of
those
reasons
we
think
we
are
going
to
a
decentralized
serverless
compute
platform,
and
that
is
effectively
the
third
step
in
the
evolution
of
compute,
which
started
off
with
mainframes
internally
hosted
moved
over.
You
know
two
decades,
three
decades
to
centralize
Cloud
platforms,
but
that
ultimately
decentralized
serverless,
compute
platforms
are
the
next
and
potentially
final
stage
of
compute,
because
they
will
be
cheaper.
They
will
be
more
resilient
and
will
offer
attributes
that
the
cloud
can't
compete
with.
B
Why
don't
we
have
this
and
the
answer
is
it's
hard
and
you
generally
have
a
sort
of
a
version
of
the
trilemma
with
decentralized
compute,
where
you
can
be
verifiable,
but
it's
hard
to
be
scalable
and
you
can
have
generalized
compute
capacity,
but
you
can't
be
verifiable
right
and
so
getting
all
three
of
these
things
solved
is
very,
very
hard,
and
so,
if
you
look
at
blockchain
verifiable
for
sure
but
scalability
in
terms
of
speed,
not
really
there
and
you
can
have
generalized
compute
capacity
like
anyone
in
in
the
cloud,
for
example-
but
it's
not
verifiable,
so
decentralization
requires
these
three
things
and
that's
what
fluence
has
been
working
on,
and
so
we
have
verifiability
through
a
variety
of
options.
B
We
can
talk
about,
and
the
real
important
thing
here
is
that
our
competition
isn't
other
platforms,
it's
not
layer
ones,
it's
not
blockchain,
it's
not
smart
contracts,
it's
AWS
and
AWS
and
Azure
and
Google
Cloud.
Are
you
know
the
massive
World
here
that
we
all
want
to
compete
against
and
that's
what
we're
we're
working
on.
So
what
does
fluence
exist
of
there's
two:
we
divide
into
two
pieces,
a
developer
platform
and
then
a
compute,
Marketplace
and
so
developer
platform.
B
Quite
quickly,
we've
rebuilt
effectively
a
decentralized
stack
and
this
each
of
these
components
there's
a
Lambda
component
which
is
Marine
and
then
like
a
rest
which
is
aqua,
which
is
effectively
the
orchestration
layer,
and
then
we
have
fluent
CLI
and
then
data
subnets,
which
basically
allow
the
deals
to
happen
that
where
compute
is
arranged.
So
that
is
the
difficult
part
here.
If
it
was
just
a
marketplace
where
I
go,
find,
someone
offering
me
Coq
capacity
and
I
make
a
deal
with
them.
B
That's
one
level
of
solution,
creating
a
software
stack
that
allows
you
to
mimic
the
cloud
in
a
decentralized
manner.
That
is
the
next
level
of
complexity
and
that's
what
we've
been
working
on
since
2017.,
compute,
Marketplace
I
think
everybody
here
understands
this,
but
this
is-
and
this
is
actually
important
piece,
because
this
is
where
the
filecoin
ecosystem
comes
into
play
and
where
we
hope
to
be
very
tightly
integrated
with
the
filecoin.
And
you
know
we
want
a
you
know,
obviously,
a
Marketplace
of
compute
providers
and
what
are
compute
providers.
Well.
B
There
are
a
lot
of
them
right
here,
because
they're
filecoin
miners
who
right
now
are
storing
data
but
have
unused
compute
capacity.
That
was
originally
that
was
used
initially
for
sealing
and
that
compute
capacity
sits
idle.
And
so
we
think
that
fluence
will
be
a
new
Revenue
stream
for
filecoin
miners,
first
to
prove
compute
capacity,
which
I'll
talk
about
and
then
to
actually
use
and
serve
that
compute
capacity
to
date
and
and
basically
compute
on
the
data
that
is
stored
by
these
Minor
by
the
miners.
And
so
this.
B
B
On
data,
a
lot
of
other
people
in
this
ecosystem
are
attacking
this
in
different
ways,
but
overall,
this
is
an
import,
though
I
think
there'll
be
ultimately
several
Solutions,
which
are
useful
depending
on
what
your
particular
use
case
is,
but
we
all
agree
that
Computing,
on
top
of
the
data
unlocks
value
and
there's,
also
excess
compute
capacity
within
the
filecoin
ecosystem,
which
can
be
incredibly
productively
deployed.
We
also
know
that
compute
demand
is
going
to
grow
dramatically.
We've
seen
the
storage
on
filecoin
grow
significantly.
B
We
expect
that
to
continue
I'm
sure
everyone
here
expects
that
to
continue
and
the
more
data
is
stored,
the
more
compute
is
going
to
be
required
and
we'll
add
more
and
more
value.
So
we
see
that
continue
and
we
think
the
monetization
of
that
compute
will
be
a
significant
Revenue
stream
going
forward,
especially
with
falcoin
plus
and
the
useful
data
which
I
think
is
eventually
taking
over
the
ecosystem.
That
data
will
is
there
to
be
work.
Computer
run
on
top
of
it
right,
and
so
then
the
question
is:
what
are
you?
B
How
is
best
to
do
that?
And
that's
where
fluence
enters
the
picture?
A
couple
things
here:
how
are
we
verifiable?
So
you
know
proof
of
Aqua
execution,
so
every
peer
along
the
way
that
is
in
aqua
is
the
orchestration
layer.
Every
peer
that
executes
code
is
there's
a
verifiable
proof
along
that
way
and
then
for
marine.
There
is
also
there's
also
verification
as
well
proof
of
processing,
so
these
are
key
components
to
what
it
is
and
I
think
I
want
to
call
attention
to
this,
because
verifiability
is
a
critical
piece
of
this.
B
If
you
don't
have
verifiability,
you're
I
think
you're
missing
a
component
that
is
the
market
will
require
and
then
frankly,
in
the
last
part,
is
proof
of
capacity,
and
this
is
something
that
we're
launching
end
of
this
year
as
you'll
see.
But
this
is
where
Falcon
miners
come
into
play
because
here
for
falcoin
miners
to
show
and
prove
they
have
capacity
we
will
and
the
protocol
will
offer
rewards.
So
it's
a
reward
based
system.
B
The
way
filecoin
started
off
offering
rewards
for
initially
proving
you
had
storage
capacity
and
then
eventually,
that's
obviously
evolve
to
verify,
deals
and
proving
and
being
paid
for
deal
verified
deals
of
real
storage
on
it
or
following
that
ecosystem
bootstrapping
model
with
fluence,
hopefully
with
the
help
of
the
ecosystem
here,
and
so
we
think
that
storage
and
compute
together
create
a
cloud
and
create
a
huge
amount
of
value,
and
so
I
think
everyone
in
this
ecosystem
agrees
that
compute
and
storage
together
are
critical
and
you
know
we're
excited
to
be
a
part
of
that
solution.
B
So
where
are
we
and
we've
started
with
this
started?
I
think
2019,
our
test,
Nets
live
the
execution
protocol,
Aqua
is
live
and
we've
raised
15
million
since
we've
begun.
Q4
is
this
is
a
big
year
we've
been
building
for
a
long
time,
and
this
is
where
a
lot
comes
to
play.
Our
Dow
is
ready
to
go
mainnet
we're
targeting
for
Q4,
along
with
verifiability
and
our
compute
economy,
and
the
compute
economy
includes
this
proof
of
compute
reward
system
which
will
flow
via
the
Dow
and
then
we're
adding
extra
languages
in
2024..
B
So
you
know
finally,
what's
the
mission
here
and
just
like
Tim
berners-lee
when
he
created
the
hyper
hyperlink,
he
didn't
really
Envision
what
that
would
lead
to,
but
he
knew
that
allowing
people
to
connect
would
lead
to
unimaginably
interesting
things
and
he
did
it
because
it
thought
it
would
spur
Innovation.
That's
what
we
think
about
compute
and
when
you
free
compute
from
clouds
and
you
make
it
trustless
and
you
allow
people
to
build
on
other
applications
without
risk
of
being
shut
off.
The
Innovation
that
can
come
from.
B
That
is
hard
to
really
imagine,
but
we
think
it
is
a
level
of
innovation
that
is
not
maybe
not
quite
the
level
of
the
internet
and
of
with
the
hypertext.
That
is
the
foundation
of
this
whole
ecosystem,
but
it's
up
there
and
so
freeing
compute
from
Cloud's
important
and
it's
actually
our
whole
mission.
I.
B
Think
in
this
ecosystem
here
is
to
do
everything
we
can
to
build
a
decentralized
cloud
and
move
away
from
from
Google
and
Amazon,
because
if
we
don't
we're
going
to
be
living
in
a
Google,
Amazon
World
in
perpetuity
and
I,
don't
think
that's
going
to
serve
us
or
anyone
else
very
well.
So
with
that,
hopefully
we
can
all
work
together
to
do
this.
Follow
us
at
underscore
influence
underscore
project
and
I'm
the
Tom
tro
websitefluence.network.
Thank
you
for
your
time
and
thank
you
protocol
labs
for
the
support.
Thank
you.
A
Great
great
thanks
for
that,
oh
quick
question
can
I
make
it
quick
Carol
I'll
hand
you
my
mic
here.
B
C
Hey
I
I'm,
just
curious
how
you?
How
are
you
doing
it?
Because
if
you
don't
have
consensus
and
you
have
a
network
of
providers,
how
are
you
making
sure
is
that
providers
are
not
tempering
this
data,
not
changing
the
results
of
the
computation,
not
stealing
the
data?
Somehow
and
like
usually
on
most
blockchains,
you
do
it
through
consensus
protocols
like.
B
A
secret
Source,
listen!
It's
it's
not
that
secret,
because
it's
open
source,
but
it
is
I,
guess
a
couple
points.
So
one
is
not
a
blockchain
and
that's
like
the
critical
foundational
premise:
it's
off
chain,
that's
where
the
proofs
come
in
and
the
proofs
of
execution
and
so
there's
a
variety
of
different
proof
levels
we
have,
depending
on
which
part
of
the
stack
we're
in
and
so
there's
a
ZK
for
some
of
it
and
then
there's
a
proof
of
execution
for
others
and
I.
B
Think
that
I
probably
would
we're
going
to
release
the
details
of
those
in
the
net
when
we
launched
when
we
release
our
next
white
paper,
which
is
probably
about
a
month
or
two
months
away,
so
I
think
I'm
gonna
have
to
wait
on
the
exact
details
on
that
until
we
really
least
the
white
paper,
because
I
don't
want
to
give
it
away
before
that.
But
that
is
obviously
a
critical
piece
of
what
we've
been
building.
It
also
sounds
like
Perpetual.