►
From YouTube: AI Chat: Responsible AI License Review
Description
Broadcasted live on Twitch -- Watch live at https://www.twitch.tv/rhyolight_
A
A
B
B
So
yeah
so
licenses
are
boring,
no
kind
of
way
to
put
it
when
you're
a
programmer
or
a
scientist.
You
know
at
some
point
you
have
to
think
about
licenses.
How
do
you
want
people
to
use
your
code
or
not
use
your
code?
How
do
you
want
people
to
attribute
the
source
to
you
or
to
the
company
that
paid
for
your
work
or
something
like
that?
B
There's
a
lot
of
questions
about
licensing
and
in
the
same
way
that
I
think
all
new
technology
that's
being
developed,
brings
with
it
hazards
and
risks
as
it's
being
developed,
because
it
allows
us
to
do
things
we
could
never
do
before
and
when
we're
in
a
position
as
humans,
you
don't
see
my
video,
isn't
it?
Are
you
guys
having
video
problems?
Hope
not.
B
One
of
the
things
you
have
to
think
about
is,
as
as
new
technology
comes,
how
is
humanity
gonna
use
it,
and
how
should
we
regulate
that
because
sometimes,
as
was
the
case
with
dynamite
and
atomic
energy,
and
all
that
there
are
and
the
internet
there's
lots
of
unforeseen
situations
that
come
about
the
revenge
effects
of
technology?
If
you
will
so.
A
B
Talk
about
and
cloning
yeah
well
there's
a
lot
of
stuff.
So
I
want
to
talk
about
this
license
that
came
to
my
attention.
A
while
back
some
I,
don't
remember
where
I
saw
it,
but
it
piqued
my
interest
enough
that
I
thought
I
should
take
some
time
and
look
over
it.
So
that's
what
I'm
gonna
do
I'm
gonna
use
this
time
to
look
over
this
license
kind
of
figure
out
what
it's
all
about.
I
have
barely
looked
at
it
yet
so
I'm
open
for
discussion.
B
If
you
guys
have
opinions,
I'm,
I'm
monitoring,
my
stream
strengths,
everything
looks
ok
because
I'm
very
wary
things
are
going
to
crash
at
any
moment.
So
anyway,
let's
let's
get
a
license
or
well.
First,
okay,
so
I'm
gonna
download
a
license.
This
is
the
main
website
empowering
developers.
So
does
this
mean
that
the
license
is
focused
towards
developers?
I,
don't
know,
thousands
of
devs
are
working
on
machine
learning
and
AI
algorithms.
B
Open
source
collaboration
is
important
challenge
to
keep
the
technology
moving
forward,
but
the
potential
for
technology
being
used
in
unintended
ways
is
a
cause
for
caution.
Releasing
algorithms
code
or
data
can
mean
that
developers
and
researchers
must
sacrifice
control
over
the
application
over
the
application
of
these
artifacts.
B
This
can
lead
to
a
moral
dilemma
and
harm
the
dissemination
of
scientific
research
if
it
is
used
irresponsibly.
This
can
have
an
impact.
This
can
impact
us
all,
so
the
rail
or
the
irresponsible
AI
licenses,
so
there's
more
than
one
license
here:
empowered
developers
to
restrict
the
use
of
their
AI
technology
in
order
to
prevent
irresponsible
and
harmful
applications.
So
this
is
to
tool
for
developers
of
AI
and
machine
learning.
To
restrict
Howard
is
used.
B
B
B
Anyway,
most
the
AI
guidelines
have
been
written
by
people
that
really
don't
know
what
they
are
regulating:
yeah,
the
ones
that
I've
read
in
blog
post
format
those
types
of
guidelines.
They
don't
make
a
lot
of
sense.
They're
there
planning
for
the
Skynet
scenario.
You
know
I
mean
that's
one
of
the
miscalculations.
I
think
you
can
make
is
just
assuming
that's
the
this
scenario.
Okay,
let's
get
a
license.
Let's
get.
B
A
B
Can
use
the
end-user
license,
even
if
you
do
not
have
existing
and
user
license
agreement
all
right,
oh
good,
so,
okay
and
I
don't
have
to
actually
download
it
and
then
there's
a
source
code
license.
We
recommend
you
include
a
reference
to
this
license
that
eats
modular
block
of
your
source
code
in
each
file,
preferably
each
at
each
code,
function,
method,
class
data
structure,
definition
and
especially
methods
that
involve
any
algorithmic
processes.
So
there
that's
sort.
B
Usually
you
see
a
copyright
header
at
the
top
of
each
code
source
file
and
the
license
is
only
mentioned
really
in
either
in
that
copyright
header
it
just
refers
to
wherever
the
license
is,
but
it's
usually
profile,
not
let's
go
to
the
let's
look
at
the
end
user
license.
First,
let
me
just
see
how
long
this
is
so
we've
got
a
license:
user
account,
company
and
user
content,
conduct
and
prohibitions,
criminal.
B
A
B
Exactly
so
that
I'm
a
little
bit
more
interested
in
the
open
source
license
in
the
end-user
license
agreement.
So
let's
look
at
that
one.
That's
a
good
point
that
we,
you
have
much
much
less
control
over
what
users
do
with
the
products
that
you
buy.
The
EULA
everyone
ignores
you
Liz
honestly.
Has
anyone
read
a
EULA
that
came
with
came
with
a
product
that
you
personally
bought
I
understand
like
companies
have
to
pay
attention
to
that
stuff,
but
generally
people
ignore
them
and
we're.
B
B
B
A
B
B
That
would
be
me
if
I
were
sending
using
this
everybody
rips.
Yes,
why
do
you?
Would
you
not
rip
the
tag
off
the
mattress
once
once
the
product
is
no
I'm,
not
gonna,
go
in
then
I'm,
not
a
libertarian,
but
I
have
a
libertarian
tendencies,
which
I
think
everyone
does
don't
tell
me
what
to
do
with
my
stuff.
I
bought
right
anyway,
anyway.
Stop
getting
me
off
track
mark,
so
I
would
be
I'm.
B
Is
this
a
decent
license
to
attach
to
the
source
code
we
we're
creating
in
the
community
or
Numenta?
You
know
that's
my
question.
What
I
want
to
get
out
of
this
is
should
I
recommend
anything
that
we,
you
start
using
this
license
so
I'm
thinking
it
from
I'm
thinking
of
this
and
almost
entirely
from
the
license
or
standpoint
but
I.
Also.
We
also
have
to
think
of
it
about
the
bit
from
the
people
who
are
going
to
use
a
software,
because
we
don't
want
to
restrict
them
too
much
or
people
won't
use
it.
B
B
Okay,
so
there's
always
a
tricky
idea
that
the
GPL
introduced,
which
is
what
does
it
mean
to
distribute
code,
but
we'll
get
to
that
point
when
we
get
to
it.
I
guess,
I
haven't
said
anything
about
that.
Yet
contribution
means
the
original
software
additions
to
the
original
software
modifications
to
the
original
software
derivative
works
of
the
original
software
contribution.
B
So
we're
also
talking
about
oh
yeah,
okay,
so
we're
talking
about
attribution,
attribution,
a
contribution
and
the
contributor
is
any
person
who
provides
a
contribution.
Okay,
additions
about
applications
are
derivative
works,
okay,
but
this
is
tricky.
Derivative
works,
derivative
works
of
the
original
software.
B
You
can't
I
can't
control
how
someone
Forks
my
code
and
I
guess
that's
the
point
of
the
license,
we're
gonna
try
to
control
and
that
you
can,
under
the
like
the
a
GPL,
for
example,
that
if
we
apply
that
license-
and
somebody
could
fork
it
and
make
a
derivative
work
which
is
new,
pick
core
I'm,
not
new
before
you
picked
that
CVP,
that
community
fork
is
a
derivative
work,
so
we
still
enforce
that
a
GPO
over
top
of
that
derivative
work.
So
that
makes
sense
now
for
most
licenses.
B
They
say
that
whoever
has
the
majority
of
the
copyright
of
the
work
and
the
derivative
work.
Then
you
know
Forks
off
of
the
original
work.
So
at
some
point
there
could
be
the
majority
of
that
work
could
now
be
in
the
hands
of
other
another
agency,
another
group
or
another
it's
out
of
our
control
once
they
have
contributed
more
to
the
original
work
than
the
original
work.
They
have
a
right
to
change
the
license.
So
that's
just
something
to
keep
in
mind.
B
B
So
well
now,
now
we're
getting
off
see,
that's
another
topic.
That's
it
look.
I
think
we've
talked
about
this
and
about
our
license
strategy
and
patent
strategy
on
our
blog
and
stuff,
but
we
encourage
hobbyists
to
make
a
better
mousetrap
if
they
publish
it
and
show
people
how
to
do
it,
absolutely
I,
don't
think
we
would.
We
wouldn't
go
after
anybody
for
doing
something
like
that.
If
they
don't
try
and
show
people
how
to
do
it
and
they
just
commercialize,
it
may
keep
their
contributions,
proprietary
and
commercialize
it
and
then
that's
the
problem.
B
So
wait
all
right.
So
we're
through
the
terms
I.
Think
I
did
there's
a
couple
things
here,
like
the
eight
section
that
does
it
doesn't
exist,
and
then
this
derivative
work
I
made
a
note
about
all
right,
so
grant
of
rights.
Subject
to
this
license,
each
contributor
grants
you
a
non-exclusive
worldwide
royalty-free
copyright
license
to
reproduce
its
contribution,
prepared
derivative
works
of
its
contribution
and
distribute
its
contribution,
or
any
derivative
works
of
its
contribution
that
you
create.
B
Okay,
let's
break
this
down
each
contributor,
so
they're
basically
saying
everyone
who
has
helped
make
this
work
is
allowing
you
all
this
royalty-free
copyright
license
to
reproduce
it
prepare
a
derivative
works.
You
could
fork
it
you
can.
You
can
take
it
and
do
something
different
with
it
and
distribute
the
contribution.
So
here's
the
first
time
that
we're
seeing
distribute
is
right
here.
So
I
did
another
thing.
B
About
is:
does
this
license
play
nice
with
the
a
GPL,
because
all
of
our
code
that
contains
HTM
intellectual
property
is
license
a
GPL.
So
if
this
says
something
counter
to
the
way
the
GPL
considers
distribution,
it
might
be
hard
to
at
work.
So
this
is
saying:
prepare
derivative
works
of
the
contribution,
so
I'm
thinking,
that's
like
creating
a
fork
and
adding
my
own
stuff
to
it
and
distribute
its
contribution.
I'm
assuming
that's
the
derivative
work.
B
B
So
I'm
wondering
if
this
distribution
part
is
a
problem
I'm
confused
about
that
I'll
have
to
think
about
this
distribution
part.
This
is
the
first
time
like
I
said:
we've
talked
about
distribution
and
I,
and
if
you
look
at
the
a
gplv3,
oh
it
has
it's
a
little
bit
more
complicated,
but
when
it
talks
about
distribution.
B
B
A
B
Written
before
there
was
all
these
software
as
a
service,
the
software,
the
service
model
existed,
so
they
didn't
think
of
they
thought
of
distribution
is
essentially
bundling
up
code
and
shipment
and
deploying
it
with
your
application,
sending
it
somewhere
right,
sending
it
somewhere.
It's
for
you
to
be
installed
on
your
computer
right
that
that
was
distribution,
but
with
the
internet
came
with
more
internet
services
came
from
I,
don't
think
it
matters
whether
it's
binary
or
whether
its
source
code
either
one
to
think
but
but
the
movement
of
that
from
point
A
to
point
B.
B
So
it
can
be
run
on
point
B.
That
was
the
distribution
now,
but
then
people
would
start
thinking.
Oh,
we
can
get
around
this
because
I
will
just
I'm
not
shipping
it
to
a
computer
I'm
just
going
to
take
the
source
code
that
I
have
it
that
when
I
have
deployed
to
a
customer,
it
doesn't
deploy.
I
just
stand
up
a
service
and
the
customers
come
to
the
service,
and
that
is
that
the
a
GPL
essentially
says
when
you
stand
up
a
service
like
that,
you
have.
B
If,
if
the
code
is
a
GPL,
then
that
is
distribution.
So
so
that's
like
we've
got,
but
you
can
get
around
that
too
I
mean
you
can
create
a
you.
Can
you
can
then
I'm
I
don't
want
to
talk
about
the
J.
A
GPL
too
too
much
I
do
need
to
make
sure
that
what
it
considers
distribution,
distribution,
distribute.
Okay,.
B
B
Sir
copyright,
okay,
where
is
there,
is
a
certain
like
freeze
in
here?
Maybe
I
was
in
the
right
place.
Maybe
it's
this
thing
about
propagation
to
propagate
a
work
means
to
do
anything
with
it
that,
without
permission,
would
make
you
directly
or
secondarily
liable
for
infringement
under
applicable
copyright
law
except
executing
it
on
a
computer
or
modifying
a
private
key,
okay,
so
propagation,
so
they're,
considering
not
like
running
a
program,
is
not
a
propagation
of
the
program.
B
B
B
This
is
why
we
we
wanted
to
use
the
8
beats
GPL,
as
because,
in
this
patent
peace
provision,
you
may
not
convey
a
covered
work
if
you
are
a
party
to
an
arrangement
with
a
third
party,
if
you,
if
you
are
a
party
to
an
arrangement
with
a
third
party,
you
are
a
party
to
an
arrangement
with
a
third
party
that
is
in
the
business
of
distributing
software.
That's
not
legalese
is
legalese.
B
Go
back
to
this
is
something
that
I
think
I'll
have
to
just
keep
in
mind
as
I
keep
going
all
right.
If
you
distribute
any
portion
of
the
contribution,
you
must
include
a
complete
copy
of
this
license
with
the
distribution.
That's
reasonable
and
you
agree
that
the
contribution
or
any
derivative
work
of
the
contribution
will
not
be
used
by
you
or
any
third
party,
subject
to
your
control.
B
Here's
I
think
the
meaty
stuff,
so
there's
two
sections
three
sections
here:
surveillance,
computer
generated
media,
healthcare,
criminal,
four
sections,
okay,
so
there's
surveillance
restrictions
will
not
be
used
to
detect
or
infer
any
legally
protected
class
or
aspect
of
any
person,
as
defined
by
the
US
federal
law,
detect
or
infer
any
legally
protected
class
or
aspect
of
any
person
as
defined
by
US
federal
law.
So
what
a
legally
protected
class
is
is
not
defined
except
by
federal
law.
What
is
this
even
I
sort
of
know?
B
B
A
B
That
doesn't
keep
it
at
that
from
happening,
I
mean
bias
by
sort
of
bias
can
exist
no
matter.
If
any
of
these
I
mean
bias
will
still
exist,
it
depends
I
mean
you
can
bias
is
something
you
can't
anticipate
in
some
aspects,
so
right,
yeah,
bad
credit
approval
based
on
extracted
properties.
That's
that's
right,
and
this
says
nothing
about
that.
You
can't
can't
prevent
that.
B
That's
all
I
mean
you
could
put
in
place
something
that
says
you
shouldn't
look
for
bias
in
these
ways
or
something,
but
this
I
don't
even
know
if
I,
so
we
so
we're
going.
We
want
to
prevent
so
basically,
this
is
saying
that
no
one
is
going
to
be
able
to
use
our
software
to
specifically
call
out
aspects
of
people.
So
what
if
I
wanted
to
find
all
the
blue-eyed
people
in
the
world
I
don't
know
and
and
I,
and
we
wanted
the
software
to
somehow
do
that
that
would
be
illegal.
That
would
be
restricted.
B
Doesn't
really
make
sense,
I
mean
this
seems
it's
basically
legally
protected
class
or
aspect.
So
that's
what
I
don't
understand
as
defined
by
you.
What
is
a
legally
protected
class
four
aspects,
the
US
federal
law,
legally
protected
class
or
aspect.
Is
this
just
somebody
making
this
up
types
of
discrimination?
B
What
are
the
protected
classes
under
federal
law,
the
EEO
terminology,
race,
religion,
color,
equal
opportunity,
equal
employment,
opportunity
program?
That's
not
what
I
want
employment
discrimination
types
of
discrimination,
so
so
is
this
the
type
of
thing
we're
looking
for
like
something
that'll
identify
age,
see:
that's
we
can't
do
this.
This
doesn't
seem
right.
I
mean
a
lot
of
these
programs
that
I
assume
these.
These
applications
that
I
expect
people
to
make
with
our
technology
is
going
to
be
to
try
and
identify
some
of
these
things.
B
What
people
do
with
that
power
is
sort
of
up
to
them.
You
could
use
that
to
make
improvements
for
all
of
these.
You
could
try
and
find
elderly
people
who
are
who
are
vulnerable
to
scammers
and
find
and
create
an
agent
that
finds
those
people
so
that
you
can
inform
them
and
educate
them
about
the
scams
right
and
yeah.
This
does
certainly
varies
by
country,
but
this
is
an
example
of
those
types
of
things
that
they're
talking
about
here.
B
B
Guess
this
is
the
the
classic
argument.
You
know
the
the
criminals
are
going
to
be
doing
this
either
way.
So
do
we
want
to
use
the
same
power
to
prevent
them
from
targeting
these
people,
or
the
unethical
people
I
should
say,
are
gonna
be
doing
this
either
way
so
I,
don't
know,
I
just
see
a
lot
of
application
from
being
able
to
identify
certain
types
of
people
in
order
to
help
them
and
I
would
hate
to
disqualify
all
applications
of
a
certain
technology
in
a
in
a
huge,
huge
area
such
as
that.
B
B
You
could
turn
on
an
agent
like
we're
I
assume
we're
going
to
have
agents
in
video
game
worlds
soon
enough.
That
I
mean
we
already
do,
but
like
intelligent
agents
and
you
could
an
agent
might
be
able
to
identify
types
of
like
stereotypes
of
players
in
the
world
just
based
on
how
they
play
the
game.
You
know,
and-
and
we
might
not
even
be
able
to
really
verbalize
with
that
with
them,
but
they
would
be
surveilling.
A
B
Classifying
on
lots
of
different
ass
of
a
person
and
whether
they're
legally
protected
or
not,
the
intelligent
thing
might
not
even
be
able
to
identify
that
at
the
time
and
it's
tricky,
you
know
you
have
to
define
two
measures,
so
researchers
would
run
afoul
of
this
clause.
You
have
to
define
to
measure
yeah.
It
seems
that
way.
It
seems
it's
like
I'm
running
into
the
very
first
thing,
I'm
like
wow,
that
that
cuts
off.
Basically
everything
we
would
do
with
an
intelligent.
A
B
I
mean,
and
we
wanted
to
interact
with
human
beings,
we
have
to
have
it
understand
human
beings
and
if
they
can
detect
or
infer
legally
protected
class
or
aspects
of
people,
how
are
they
going
to
be
intelligent?
Any
human
has
the
ability
to
do
this
to
detector
and
fur
illegally,
protect
the
class
or
aspect
of
any
person
and
we're
seeing
right
off
the
bat
number
one.
You
can't
do
that
that
that
crosses
intelligent
things
off
to
the
chart
right
that
I.
B
B
A
B
Won't
work
you
just
have
to
try
and
guide
it,
and
if
you
guide
it
to
restrictive
lis,
it's
gonna
everybody's
gonna
be
like
nope
whatever,
and
then
oh
you're,
trying
to
guide
all
the
good
guys
to
do.
One
thing:
well,
while
the
technology
wants
to
have
is
showing
an
explosion
in
another
field,
you
have
to
follow
the
technology,
and
this
this
doesn't
seem
to
be
following
what
I
would
consider
an
explosion
of
intelligent
things
or
like
the
creation
of
a
GI.
B
Mark
says
this
is
what
I
opened
with
they
are
trying
to
do
the
right
thing,
but
are
not
understanding
what
they're
saying
means
yeah,
it's
I
think
that's
the
same
thing.
They
have
to
know
that
it's
a
bad
thing
to
know
how
to
not
do
it.
They
have
to
know
what
it
is
a
bad
thing
to
know
how
to
not
do
it
right,
yeah.
B
All
right,
so,
let's,
let's
move
on
because
so
this
is
a
total
non-starter
for
me
again.
I
think
we
sort
of
agree.
Oh,
let's
get
some
another
comment
from
Peter
Kovalev.
Is
it
possible
to
kind
of
serialize
HTM
sequence
memory
as
a
huge
tree
of
binary,
St
eras
over
time
snapshots
of
binary
activity
of
all
the
neurons
of
the
columns?
That's
certainly
possible
yeah.
You
can
definitely
do.
A
B
A
B
Like
I
said
it's
leaving
sort
of
a
trail
of
synapse
strengths
as
its
as
its
feeding
data
into
it.
Yeah
hey,
thank
you
for
the
follow
single
mingle.
This
is
not
a
singles
chat,
but
I'm
glad
you're
here,
ACOs
already
apologizing.
It
seems
not
to
be
possible
to
lower
the
stream
quality
at
this
time.
I
have
no
control
over
the
stream
quality.
It's
it's
twitch
is
supposed
to
decide
what
stream
quality
you
should
get
and
deliver
it
to
you
and
I'm.
Sorry,
if
you're
having
a
bad
twitch
experience.
B
Computer-Generated
media-
let's
see
what
this
says:
synthesized!
Oh
it's
another
thing:
you're
not
supposed
to
do.
Synthesized
and/or,
modify
audio,
realistic
and
or
video
realistic
representations,
indistinguishable
from
photo
video
recordings
of
people
and
events
without
including
a
caption
watermark
and/or
metadata
file
indicating
that
the
audio,
realistic
and
or
video
realistic
representations
were
generated.
Using
the
contribution,
I
think.
A
B
Probably
get
on
board
with
that,
but
I
mean
you're
gonna
be
constantly
having
that
logo
everywhere,
I
mean
eventually,
people
are
going
to
just
ignore
it,
because
it's
going
to
be
everywhere
so
again,
I,
don't
think
it's
really
gonna
help.
I
think
that's
something
that
I
mean
we're
creating
our
own
reality
here
and
everybody
creates
their
own
reality.
B
We're
gonna
have
to
use
our
brains
to
identify.
What's
real
and
what's
not
just
like,
we
always
have
that's
just
the
world
that
we
live
in
so
but
that's
not
awful
I
think
you
could
work
around
that.
Every
every
piece
of
art
is
gonna
have
to
have
a
big
stamp
on
it.
That
says
this
was
not
created
by
a
human
or
had
contributions
from
the
robot
I,
don't
know,
but
I,
don't
know,
I
mean
think
about
all
of
the
music
that
you
hear.
That's
somewhat
you.
B
Is
happening
you
know
so
I
wouldn't
say
that's
a
non-starter,
but
I'd
say
that
I,
don't
think
any
I,
don't
think
I,
don't
think
it's
a
good
I,
don't
think
it
matters.
I
think
people
are
gonna
believe
what
they
want
to
believe,
whether
there's
a
thing
on
there
or
not.
And
then
you
can
just
start
taking
other
videos
and
stamping
the
thing
on
there
and
saying
here's
Tom
Cruise
on
Oprah,
but.
B
B
B
Yeah
exactly
mark
I,
just
my
mom
and
grandma
were
just
in
town
and
and
I.
Have
this
TV
stand
I
have
a
TV
stand
like
the
kind
you
just
pulled
out
right
and
I
said
and
I
was
telling
my
mom
how
ridiculous
prop
65
has
made
these
warnings
in
California
and
I?
Look
at
the
TV
stand
on
the
bottom
of
it,
and
it
says
the
TV
stand
is
known
by
the
state
in
California
to
cause
cancer.
The
TV
stand.
B
B
I
mean
I'm
worried
about
the
health.
The
the
thing
is,
the
companies
need
to
be
held
accountable,
the
algorithms
they
use
aren't
the
problem.
It's
it's
the
companies
and
the
policies.
It's
not
the
algorithms
that
they're
using
predict
the
likelihood
anyone
any
person
to
final
insurance
plan
based
on
turning
lifestyle
over.
That's
again,
that's
that's
what
humans
currently
do
and
I
mean
you
can't
tell
the
AI.
You
can't
restrict
AI
applications
to
not
do
what
we're
already
doing.
B
I,
don't
see
what
the
point
of
that
is,
determine
an
insurance
premium
or
deny
insurance
applications
or
claims
based
on
data
determining
a
lifestyle
of
the
person,
medical
test
reports,
demographic
details
of
a
person.
I,
don't
have
a
problem
with
any
of
these.
As
long
as
you
is,
you
restrict
the
company
and
the
humans
to
not
the
algorithms
I
mean
if
you
want
to
prevent
healthcare
insurance
companies
from
predicting
the
likelihood
that
any
person
will
request
a
filed
insurance
claim
make
that
the
law
don't
restrict
it
to
the
software.
B
Warnings
for
lawsuits,
I
I
mean
I,
get
it
I,
just
it's
just
silly
who's
gonna.
How
can
I
prove
to
anybody
that
my
TV
table
caused
grandmas
cancer?
So
so
these
all
seem
I'm
not
saying
these
are
bad
ideas.
I'm
just
saying
it's,
it
doesn't
seem.
You
should
apply
it
to
the
situation,
not
the
algorithm,
diagnose
a
medical
condition
without
human
oversight.
B
So
that's
sort
of
interesting
but
I
still
think
I
mean
that's
what
we
want
right.
I
mean
again.
This
should
be
an
application.
This
restriction
should
be
applied
to
an
insurance
company
or
a
doctor's
office,
not
the
software,
of
course.
We're
gonna
want
software
that
diagnoses
a
medical
condition.
That's
what
we're
trying
to
find.
We
want
an
algorithm
that
says
this
person
most
likely
has
cancer
or
will
get
cancer
in
ten
years
or
something
we
need.
We
if
we
could
find
that
that'd
be
amazing
right.
If,
if
we
apply
this
license,
we
can't
do
it.
B
So
that
I
mean
we're
talking
about
like
a
Minority,
Report
style
right
by
any
person.
Does
that
mean
any
one
person?
Does
that
mean
like
analyzing
Fred
Johnson
in
Butte,
Montana
and
like
feeding
and
having
an
agent
analyze
everything
about
that
person
and
and
then
predict
whether
when
he's
gonna
commit
a
crime
or
something
like
that,
the
likelihood
that
this
person's
gonna
commit
a
crime
I
could
see
that
yeah
I
can
see
that
I
wouldn't
want
an
algorithm
doing
that.
B
Unless
there
is,
you
know
this
person
is
under
surveillance
because
you
know
they're
criminal,
you
know,
or
they
have
there's
some
evidence
against
them,
that
they
have
or
will
commit
a
crime.
But
if
this
is
just
just
an
umbrella
thing
like
nobody
can
just
pick
a
random
person
and
do
take
do
the
work
it
takes
like
even
if
you're
a
detective,
you
know
nobody
would
allow
you
to
pick
Fred
Johnson
in
Butte
Montana
and
just
spend
the
next
week
of
your
life
analyzing.
Whether
he's
gonna
predict
a
crime
or
not.
B
You
know
we
don't
have
a
system
set
up
to
do
that,
so
I
get
the
point
they're
making
they
don't
want.
The
AIS
were
creating
to
try
and
do
this
on
everybody
yeah.
It's
the
precogs
idea,
so
I
get
that
but
again
make
this
law
an
organizational
law,
not
specifically
about
the
software
that
we're
putting
because
I
might.
B
We
might
want
to
have
a
piece
of
software
that
has
the
ability
to
analyze
a
person,
especially
if
they
are
a
criminal
if
they
have
a
record
and
we
as
and
the
police,
already
suspect
them
of
criminal
activity.
I
mean
that's
a
tool
in
the
crime,
justice
toolbox,
right
particular
likelihood
of
any
person
being
a
criminal
or
having
committed
the
crime.
It's
the
same
thing
predict
the
likelihood
any
person
being
criminal
based
on
the
person's
facial
attributes
or
another
person's
or
another
person's
facial
attributes.
B
B
B
A
B
This
good,
but
this
is
again
these
are
these-
are
policies
that
should
be
applied
to
humans
and
machines.
Not
just
humans
are
not
just
machines.
It's
not
about
the
software
that
we're
writing.
These
restrictions
are
higher
level
things
and
we
need
everyone
needs
to
be
aware
of
these
things,
not
just
the
people
writing
AI,.
B
Name,
family
name
yeah,
so
this
is
all
just
about
anti-discrimination
for
criminal
stuff,
heart
rate,
pest
break,
heart
rate,
respiration,
breathing,
so
they're
covering
all
their
all
the
bases
here.
So
they
don't.
They
don't
want
to
use
any
of
those
types
of
data
to
predict
the
likelihood
of
a
person
being
a
criminal
based
on
the
person
or
other
users
or
other
users.
B
B
B
B
That
working
out
the
bias
is
in
the
data
I
mean
the
the
data
bias
is
going
to
be
a
problem,
even
if
you,
even
if
you
restrict
all
this
in
software,
the
data
bias
is
going
to
come
out
when
the
human
analysis,
so
the
mark
says
stress
in
people
going
through
an
airport
or
bank
line,
it's
in
the
face
in
the
body
yeah.
Absolutely
that
doesn't.
B
Different
people
and
different
cultures
express
different
emotions
in
different
ways
and
some
cultures.
Some
emotions-
don't
even
exist.
So
it's
like
emotion
is
more
of
a
cultural
artifact
than
an
evolutionary
artifact.
Well,
a
genetic
artifact,
I
would
say,
I
think
you
can
only
use
this
information
under
special
provision
from
the
authorities.
Well
yeah.
But
this
says
this
is
basically.
B
Okay,
I'm
not
terminating
the
license,
so
I
already
sort
of
know
that
this
isn't
going
to
work
for
for
any
real
AGI
type
of
software.
It
might
make
sense
for
some.
You
know
we
gay
I,
existing
we
ki
stuff,
but
again
the
data
biases
is
not
a
really
addressed
here,
except
I
did
the
data
bias
is
addressed
here
by
basically
restricting
any
prediction
of
criminal
activity
at
all
based
on
any
data,
so
you
sort
of
admit
the
bias
when
you
emit
the
data,
so
they
fix
that
problem.
B
B
Right,
maybe
there's
some
comment
from
this
on
some
other
AI
forum.
Somewhere,
people
must
have
discussed
this
before
there
you
go,
can
maybe
you
guys
could
could
search
that
look,
look
up,
responsible,
AI
license
and
see
if
there's
any
other
opinions
on
it,
the
problem,
I
sort
of
know
what
the
problem
is
limited
control
or
how
it's
used.
B
B
B
Moral
dilemma:
why
how
real
works?
Restricting
AI
ml
software
Muses
to
the
list
of
harmful
applications,
surveillance
for
contradiction,
allowing
other
applications
key
goal
of
real
teams
to
advance
in
import
tint
conversation?
How
can
we
move
from
mere
talk
to
real
action
when
it
comes
to
developing
responsible,
AI
and
m/l
technologies?
The
real
team
believes
this
will
require
significant
changes
in
the
practice
of
our
profession,
including
rethink
fundamental
research
like
software
licenses.
B
We
expect
this
will
be
controversial
but
anticipate
that
the
ensuing
discussion
will
be
highly
constructive,
Levy's
new
versions
of
realizes
feedback
via
our
website.
So
here's
where
I'll
give
feedback?
Oh.
This
is
just
the
website
I'm
on,
so
you
know
what
I'll
do
I'll
just
and
I'm
done
with
this
I'll
just
post
the
whole
stream
they
can
watch
in
an
hour
while
Ison's
has
already
know
why
licenses.
So
they
only
have
the
two.
Oh
there's
a
team,
this
person,
so
here's
Daniel
Macduff.
B
This
person
is
called
Danish
contractor
it's
kind
of
weird
Julia,
Haines
Danish
contractor
and
he
doesn't
look
Danish.
It
could
be
a
contractor,
but
Danish
contractor
is
an
AI
researcher,
but
he's
not
a
character.
I,
like
that
name,
Christopher
Heinz,
he
that
reminds
me
he
brought
Houston's
familiar
and
Brent
Hecht
I,
don't
know
those
people
off
top
my
head,
whose
definition
of
responsible
good
question.
B
Why
are
you
focusing
only
on
for
domain
areas?
It's
the
first
version
of
the
license.
Okay,
so
maybe
they'll
add
weapons,
violence
or
something.
Why
are
you
focusing
only
on
AI
applications?
Good
question
the
same.
That
means
the
same
algorithm
trained
on
different
data
can
serve
a
very
different
purpose
than
it.
No
it
doesn't.
It
doesn't.
B
B
B
You
mean
regulating
those
constituent,
you
mean
you
want
to
regulate,
GPU,
accelerate
acceleration
in
quantum
computing
I'm,
not
sure
I'm,
not
sure
why
you
want
to
regulate
GPU
acceleration
Jeremy
says:
do
you
think
this
license
is
just
to
protect
the
developer
from
being
held
responsible,
I
mean
say
someone
did
take
your
code
that
was
released
under
this
license
and
used
it
for
nefarious
outcomes.
What
recourse
would
developer
even
have
to
stop
it?
You
could
sue
that's
about
it,
but
so
so
yeah
you
could.
You
could
sue.
B
Because
they're
violating
the
license-
and
that
is
expensive,
so
you'd
have
to
have
a
really
good
reason
to
do
it,
so
they
have
to
be
really
violating
the
license.
I
would
think,
but
I
think
that's
all
you
could
do
and
that's
if
it's
in
the
United
States,
you
know
if
it's
international,
it's
a
whole
other
thing
just
go
with
the
three
loss
of
robotics
and
be
done
with
it.
B
That's
not
so
I
mean
that's
not
a
bad
idea.
I
like
the
Three
Laws
of
Robotics.
Alright,
so
that's
that's
it
everybody!
We
talked
about
the
responsible,
AI
license
and
sort
of
decided
that
at
this
point
in
time
it
was
not
something
we
were
going
to
consider
and
if
you
have
any
questions
or
comments
about
that,
come
to
HTM
forum,
and
you
can
comment
there.
B
If
I
can
I
I
messed
up
my
here's,
here's
the
thing:
I
messed
up,
my
chat
bought
a
little
bit,
but
there's
the
like
taste
uniform,
hope
that
works
and
I'll
put
this
video
on
YouTube.
So
thanks
for
watching,
if
you're
watching
on
YouTube,
don't
forget
to
subscribe
to
the
channel
and
like
the
video
thanks
a
lot
and
if
you're
watching
on
twitch,
follow
me
and
check
out
my
events
page
on
my
twitch
channel
I've
got
it
laid
out
for
the
rest
of
the
week.
I'm
gonna
try
and
keep
it
a
week
ahead.