Play video to start
Duration: 2m07s
Tap 
to scroll to a caption.
Tap word for more information.
Now, there's a frankly terrifying
Now
there's
a
frankly
terrifying
prediction from an AI expert of the risk
prediction
from
an
AI
expert
of
the
risk
to humanity from artificial
to
humanity
from
artificial
intelligence. A top safety researcher at
intelligence
A
top
safety
researcher
at
Anthropic has suggested that there's a
Anthropic
has
suggested
that
there's
a
greater than 10% chance that AI could
greater
than
10
chance
that
AI
could
kill all human beings within the next
kill
all
human
beings
within
the
next
decade. As cyber correspondent Joe Tidy
decade
As
cyber
correspondent
Joe
Tidy
is here with me. Well, Joe, I mean, this
is
here
with
me
Well
Joe
I
mean
this
is bleak in the extreme. How seriously
is
bleak
in
the
extreme
How
seriously
should we take this? Well, as you say,
should
we
take
this
Well
as
you
say
it all started with these posts on the
it
all
started
with
these
posts
on
the
social network X, formerly Twitter, by
social
network
X
formerly
Twitter
by
two people. Uh, the first was an AI
two
people
Uh
the
first
was
an
AI
researcher. Used to work for OpenAI, the
researcher
Used
to
work
for
OpenAI
the
makers of Chat GPT, now works for
makers
of
Chat
GPT
now
works
for
Anthropic, the makers of Claude, but
Anthropic
the
makers
of
Claude
but
this morning resigned. Uh, this is what
this
morning
resigned
Uh
this
is
what
Jacob Coxen posted. He said that the
Jacob
Coxen
posted
He
said
that
the
companies are acting irresponsibly on AI
companies
are
acting
irresponsibly
on
AI
safety and gambling with our lives. It's
safety
and
gambling
with
our
lives
It's
a pretty stark warning. But then perhaps
a
pretty
stark
warning
But
then
perhaps
even more arresting came the reply, as
even
more
arresting
came
the
reply
as
you say, from this uh very high-profile
you
say
from
this
uh
very
high-profile
uh head of alignment at Anthropic
uh
head
of
alignment
at
Anthropic
current head of alignment as well who
current
head
of
alignment
as
well
who
frankly and and bluntly agreed with that
frankly
and
and
bluntly
agreed
with
that
uh proclamation. Evan Hubinger said the
uh
proclamation
Evan
Hubinger
said
the
company earnestly believe AI could kill
company
earnestly
believe
AI
could
kill
all humans. He cheerily added the
all
humans
He
cheerily
added
the
exclamation mark there as well. And he
exclamation
mark
there
as
well
And
he
said that I personally think there is a
said
that
I
personally
think
there
is
a
more than 10% chance of it happening in
more
than
10
chance
of
it
happening
in
the next 10 years. Well, the comments as
the
next
10
years
Well
the
comments
as
you can imagine have shocked people
you
can
imagine
have
shocked
people
around the world. There's been
around
the
world
There's
been
discussions even in parliament today
discussions
even
in
parliament
today
about potential need for some sort of
about
potential
need
for
some
sort
of
mandated slowdown of AI development. But
mandated
slowdown
of
AI
development
But
these apocalyptic predictions
these
apocalyptic
predictions
are not new. We've heard this kind of
are
not
new
We've
heard
this
kind
of
thing before. But I think the reason why
thing
before
But
I
think
the
reason
why
people are so focused on it now, the
people
are
so
focused
on
it
now
the
reason why now and today people are
reason
why
now
and
today
people
are
especially worried is because of the
especially
worried
is
because
of
the
summer that we've had, there have been
summer
that
we've
had
there
have
been
scandal after scandal where companies
scandal
after
scandal
where
companies
like Anthropic and OpenAI and Meta have
like
Anthropic
and
OpenAI
and
Meta
have
revealed that AI agents that should have
revealed
that
AI
agents
that
should
have
been contained and safe have broken out
been
contained
and
safe
have
broken
out
of their safety systems and gone on
of
their
safety
systems
and
gone
on
uncontrollable hacking sprees. The
uncontrollable
hacking
sprees
The
companies themselves say they're
companies
themselves
say
they're
spending a lot of money, time, expertise
spending
a
lot
of
money
time
expertise
on making sure the models are safe. But
on
making
sure
the
models
are
safe
But
of course there's senior leaders now
of
course
there's
senior
leaders
now
asking for lawmakers particularly in the
asking
for
lawmakers
particularly
in
the
US to step in. Whether or not that will
US
to
step
in
Whether
or
not
that
will
happen, we don't know. Donald Trump has
happen
we
don't
know
Donald
Trump
has
recently called the AI industry a golden
recently
called
the
AI
industry
a
golden
goose.
goose
Dictionary
Tap ? for definitions, examples, pronunciation, and comprehension questions.
- Pause
0.5x - Slower
0.75x - Slow
1.0x - Normal
You might also like these videos
If you enjoy using our custom video player, why not join thousands of subscribers who have signed up to our free weekly lessons?