
so hello everybody there is a great
chance that in 50 years from now robots
will have taken over the planet and that
not that’s not necessarily a bad thing
because it can mean two things either
that robots are very very widely
accessible product or in terms of the
singularity
well the droplets will have acquired
human-like intelligence if not surpassed
it the thing is like as humans we are
used to being the most the most
intelligent species on the planet we
just don’t know what’s going to happen
if robot intelligence is going to
surpass human intelligence everything is
unpredictable so what can happen is the
other doomsday scenario with robots
killing everybody or robots turning
everybody into paperclips and don’t ask
why just because they’re more
intelligent in us they can just do
anything without justifying it so how
possible is this well there’s a lot of
debate on the subject some people say
that it’s very possible some people say
that it’s just a doomsday scenario well
my personal view is that if AI keeps
perfecting itself making herself better
and better and better the Navy in the
future we’re going to have machines as
intelligence as human if not better but
for the moment we are at this stage
so the question is how come from that
stage to the doomsday scenario of
rabbits being even more and more perfect
well the thing is that when we fire dart
that dart has a certain goal it reaches
a certain place when we create a Robert
the robot has to follow some certain
commands and it has to do a certain
function in this the case of the
self-driving car well the function of
the Robert is just to drive more and
more efficiently but in order to avoid
problems that like the ones that we saw
in the video it makes sense that we want
the robot to become more and more
perfect but my question to you is when
does this perfectionism actually stop in
the case of the self-driving car one of
its goals will be for example to never
pass with a red light to always cross
with a green light but if this robot
tries to make itself more better and
better into achieving this goal well
this would mean that it could
potentially hack into the traffic light
make all red lights green and then pass
because it does what it’s supposed to do
right it’s supposed to pass with red
light with green light and this could
possibly in in this in the in the case
of super intelligent machines these
machines by trying to make themselves
better and better and better it could
create an intelligence explosion which
would create a singularity scenario the
one that I mentioned before so how could
we possibly as humans how could we
possibly prevent this from happening
well one way would be to make Roberts
read our minds as I said if we give the
self-driving car enough context to make
it learn that for example turning all
red lights great into green just in
order to be able to pass is bad then for
example it wouldn’t really do it but the
problem with reading minds and context
is actually very very you know it’s very
very vague because human life is
actually pretty unpredictable so let’s
assume that you have a firefighter that
tries to save a baby from the building
that’s burning and you have an AI that
is supposed to help the firefighter with
this so the AI knows everything about
humans
knows everything about fires and it
knows everything about babies but do we
know that the a can actually combine
everything the AI sees the firefighter
running towards the building and he gets
the social cue of running but is the
firefighter running towards the fire
because she wants to be warm is he
running towards the fire because she
wants to save the baby or is he running
towards the fire just because he wants
to jog this kind of context is just
we’re not sure if this would be
accessible in a brain that is artificial
it really created and doesn’t have the
human gut feeling so if we cannot teach
a machine that sometimes it’s bad to be
perfect and if we cannot put enough
context on it to make it understand that
perfection can actually go very wrong
for humans the nut can be a very bad
scenario so I guess the goal for the
future would be to have an AI and teach
it that it’s that its goals have to
align at some point with human goals and
that is the hardest way to achieve it’s
the hardest thing to do because I view
us as cavemen rediscovering fire and
this fire is AI so it can go very very
well and just create make everything new
as they did with caveman discovering
fire or it can burn everything down so
it is all up to us it is all about
depends on how alert would stay and how
well we design our robots but that’s of
course depending on the manufacturers to
do so yeah that’s me thank you very much
chance that in 50 years from now robots
will have taken over the planet and that
not that’s not necessarily a bad thing
because it can mean two things either
that robots are very very widely
accessible product or in terms of the
singularity
well the droplets will have acquired
human-like intelligence if not surpassed
it the thing is like as humans we are
used to being the most the most
intelligent species on the planet we
just don’t know what’s going to happen
if robot intelligence is going to
surpass human intelligence everything is
unpredictable so what can happen is the
other doomsday scenario with robots
killing everybody or robots turning
everybody into paperclips and don’t ask
why just because they’re more
intelligent in us they can just do
anything without justifying it so how
possible is this well there’s a lot of
debate on the subject some people say
that it’s very possible some people say
that it’s just a doomsday scenario well
my personal view is that if AI keeps
perfecting itself making herself better
and better and better the Navy in the
future we’re going to have machines as
intelligence as human if not better but
for the moment we are at this stage
so the question is how come from that
stage to the doomsday scenario of
rabbits being even more and more perfect
well the thing is that when we fire dart
that dart has a certain goal it reaches
a certain place when we create a Robert
the robot has to follow some certain
commands and it has to do a certain
function in this the case of the
self-driving car well the function of
the Robert is just to drive more and
more efficiently but in order to avoid
problems that like the ones that we saw
in the video it makes sense that we want
the robot to become more and more
perfect but my question to you is when
does this perfectionism actually stop in
the case of the self-driving car one of
its goals will be for example to never
pass with a red light to always cross
with a green light but if this robot
tries to make itself more better and
better into achieving this goal well
this would mean that it could
potentially hack into the traffic light
make all red lights green and then pass
because it does what it’s supposed to do
right it’s supposed to pass with red
light with green light and this could
possibly in in this in the in the case
of super intelligent machines these
machines by trying to make themselves
better and better and better it could
create an intelligence explosion which
would create a singularity scenario the
one that I mentioned before so how could
we possibly as humans how could we
possibly prevent this from happening
well one way would be to make Roberts
read our minds as I said if we give the
self-driving car enough context to make
it learn that for example turning all
red lights great into green just in
order to be able to pass is bad then for
example it wouldn’t really do it but the
problem with reading minds and context
is actually very very you know it’s very
very vague because human life is
actually pretty unpredictable so let’s
assume that you have a firefighter that
tries to save a baby from the building
that’s burning and you have an AI that
is supposed to help the firefighter with
this so the AI knows everything about
humans
knows everything about fires and it
knows everything about babies but do we
know that the a can actually combine
everything the AI sees the firefighter
running towards the building and he gets
the social cue of running but is the
firefighter running towards the fire
because she wants to be warm is he
running towards the fire because she
wants to save the baby or is he running
towards the fire just because he wants
to jog this kind of context is just
we’re not sure if this would be
accessible in a brain that is artificial
it really created and doesn’t have the
human gut feeling so if we cannot teach
a machine that sometimes it’s bad to be
perfect and if we cannot put enough
context on it to make it understand that
perfection can actually go very wrong
for humans the nut can be a very bad
scenario so I guess the goal for the
future would be to have an AI and teach
it that it’s that its goals have to
align at some point with human goals and
that is the hardest way to achieve it’s
the hardest thing to do because I view
us as cavemen rediscovering fire and
this fire is AI so it can go very very
well and just create make everything new
as they did with caveman discovering
fire or it can burn everything down so
it is all up to us it is all about
depends on how alert would stay and how
well we design our robots but that’s of
course depending on the manufacturers to
do so yeah that’s me thank you very much
Please follow and like us: