►
From YouTube: From Quality Metrics to Reputation - Michael Zargham
Description
For more information on Filecoin
- visit the project website: https://filecoin.io/
- or follow Filecoin on Twitter: https://twitter.com/Filecoin
Get Filecoin community news and announcements in your inbox, monthly: http://eepurl.com/gbfn1n
A
Thank
you
very
much
for
having
me
so
I'm
going
to
probably
bring
us
up
stack
a
little
bit.
Not
only
does
the
filocoin
network
store
a
large
amount
of
data,
but
it
actually
produces
a
large
amount
of
data
and
we
talked
a
bit
about
all
of
that
highly
heterogeneous,
distributed
data
about
operators
and
markets
etc.
That's
available
and
in
this
talk
we're
going
to
briefly
look
at
how
we
can
put
that
data
to
work
to
make
a
sort
of
richer,
healthier
ecosystem.
So
with
that
we're
going
to
talk
about
quality,
metrics
and
reputation.
A
So,
just
keep
in
mind
that
when
we
go
to
put
data
to
work
to
actually
make
decisions,
there's
a
lot
of
processes
that
happen
after
you
get
the
raw
data.
So
we
saw
an
example
of
how
various
tools
will
actually
collect
this
highly
rich
low-level
data,
and
we
want
to
start
to
think
about
how
we
can
take
it
to
the
level
where
we
can
interpret
and
make
actions.
A
And
when
I
say
we,
I
don't
mean
file
coin
itself,
I
actually
mean
the
participants
in
the
ecosystem
and
what's
tricky
about
this,
is
in
your
sort
of
traditional
sort
of
web
2
or
corporate
setting.
There
are
highly
specialized
teams
that
develop
and
maintain
the
data
infrastructure,
especially
when
it's
the
scale
of
the
infrastructure
used
to
monitor
data
coming
off
of
the
filecoin
network.
A
So
this
particular
diagram
is
a
flow
created
by
one
of
the
block
science,
researchers,
danilo
he's
been
working
on
a
bunch
of
file
coin
metric
research
with
us
and
what
we're
seeing
is
the
way
the
individual
metrics
in
the
middle,
and
these
are
examples,
would
feed
into
various
use
cases
or
decision
making
problems
that
are
related
to
the
operation
and
consumption
of
the
the
services
on
the
file
coin
network,
whether
those
are
storage,
clients
buying
service
contracts
with
miners
or
whether
these
are
you
know,
lenders
or
insurers
in
the
emerging
file
coin
markets
to
help
cover
the
the
collateral
aspects.
A
So
we
kind
of
highlight
at
least
a
few
different
decision-making
problems
that
we're
going
to
be
expecting
participants
in
this
ecosystem.
To
experience
that
we
should
see
being
made
in
a
data-driven
way,
so
lenders
are
making
decision
about
how
much
to
lend
based
on
maybe
the
deal
verification
status
of
miners
who
they
would
be
lending
filecoin
to.
We
can
imagine
the
emergence
of
underwriters
who
actually
underwrite
those
loans
to
make
sure
that
these
lending
providers
actually
have
their
risk
managed
properly.
A
We
would
expect
all
of
these
use
cases
to
emerge
and
actually
to
emerge
individually
and
there's
a
sort
of
interesting
phenomena
where
no
one
party
owns
the
data
infrastructure
at
the
data
storage
level
or
even
at
the
sort
of
data
interpretation
level,
and
so
what
we
hope
to
see
is
a
growing
provider
basis
for
these
types
of
analytics.
A
One
example
is
a
reputation
score
and
we
put
a
lot
of
effort
into
making
sure
that
reputation
isn't
sort
of
a
god's
eye.
You
know
this
is
your
credit
score
and
if
you
do
badly
you're
doomed,
because
that
can
be
a
little
bit
overly
deterministic
and
it
can
cause
someone
who
has
a
bad
score,
maybe
to
never
be
able
to
do
anything
again,
but
in
this
case
we're
looking
at
the
scores
as
something
that
changes
over
time
and
that
which
scores
matter
might
depend
on
the
use
case.
A
So
in
this
simple
example,
where
alice
is
a
service
provider
and
bob
is
a
service,
consumer
bob
might
actually
care
about
different
quality
metrics
depending
on
what
that
service
is.
And
so,
if
this
is
a
a
consumer
of
basically
a
client
who's
going
to
consume
storage,
maybe
they
they
in
that
storage
is
maybe
small.
They
care
mostly
about
the
retrieval
time
and
the
likelihood
of
their
deal
termination
in
the
uptime.
A
So
they
weight
those
things
accordingly
and
we
can
actually
look
at
a
mixture
of
a
set
of
quality,
metrics
and
part
of
the
reason
this
is
so
important
is
because,
as
an
architecture,
it's
like
a
little
bit
more
intersubjective
yeah.
Okay,
we
have
some
low
level
facts,
but
the
way
that
we
aggregate
them
up
to
make
a
decision
might
depend
a
lot
of
the
on
the
context
of
that
decision.
A
These
various
quality
metrics,
such
as
a
reliability
measure
that
might
have
some
nuance
into
how
we
aggregate
information
over
space
and
time,
but
as
long
as
we
have
a
well-defined
and
sort
of
even
provable
reference
for
how
we
built
the
metric,
people
can
decide
for
themselves
whether
or
not
they
want
to
use
that
metric.
As
part
of
the
reputation
score
that
they
use
to
make
decisions
about
their
operations
within
the
network,
and
so
we'll
want
to
wrap
up
by
looking
at
all
ways,
some
of
these
things
could
fit
together.
A
This
is
really
what's
going
to
drive
end
use
by
people
who
don't
necessarily
need
to
understand
the
low-level
details
about
the
filecoin
network,
but
they
can
still
trust
it
to
provide
them
services,
they
make
decisions
to
use
and
participate
in
that
network
and
that
drives
outcomes.
You
know,
new
new
observations
are
made
in
the
low-level
data.
It
finds
its
way
back
up
to
the
top
again,
and
so
we
have
an
inherently
iterative
system,
even
when
everyone
is
not
trying
to
fulfill
all
the
parts.
A
So
I
invite
the
people
in
this
this
summit
to
really
think
about
where
they
can
best
provide
or
participate
in
this
ecosystem.
So
thank
you
very
much.