Do LLMs Understand? AI Pioneer Yann LeCun Spars with DeepMind’s Adam Brown. - part 1/15
2025-12-12_17-05 • 1h 15m 39s
Janna Levin (Professor of Physics and Astronomy)
00:14.640
It's
a
pleasure
to
have
Adam,
my
colleague
and
friend
and
Jan,
he's
been
with
us
before.
Jan,
you
really
are
all
over
the
news
right
now.
Janna Levin (Professor of Physics and Astronomy)
00:26.240
I've
gotten
so
many
people
forwarding
articles
about
you
this
week.
it
all
kicked
off
on
Wednesday.
Do
you
want
to
discuss
the
I
can
just
say
the
headline.
The
headline
was
the
equivalent
of
Jan
Lakun,
chief
scientist
leaves
meta.
Um,
do
you
care
to
comment?
Yann LeCun (Chief AI Scientist)
00:45.880
I
can
neither
confirm
nor
deny.
Janna Levin (Professor of Physics and Astronomy)
00:48.480
Okay.
So
all
of
the
press
that
the
core
that's
here
to
get
the
scoop,
cannot
get
the
scoop
tonight.
All
right,
well,
um
you
can
come
afterwards
and
buy
you
on
a
drink
and
see
how
far
you
get.
Yann LeCun (Chief AI Scientist)
01:03.880
I
already
had
one
but
that
was
a
Janna Levin (Professor of Physics and Astronomy)
01:07.480
The
Frenchman
had
some
wine
upstairs.
So
we
have
this
era
where
every
time
any
of
us
turn
on
the
news,
look
at
the
computer,
read
the
paper,
we're
confronted
with
conversations
about
the
societal
implications
of
AI
and
whether
it's
about
economic
upheaval
or
the
potential
for
Janna Levin (Professor of Physics and Astronomy)
01:25.880
political
manipulation
or
AI
psychosis.
There's
a
lot
of
pundits
out
there
are
discussing
this.
And
Janna Levin (Professor of Physics and Astronomy)
01:31.120
I
And
And
I
It
is
a
very
important
issue.
I
kind
of
want
to
push
that
towards
the
end
of
our
conversation
because
what
a
lot
of
people
who
are
discussing
this
don't
have
is
the
technical
expertise
that's
on
this
stage.
And
so
I
really
want
to
begin
by
grounding
this
in
that
Janna Levin (Professor of Physics and Astronomy)
01:47.520
technical
scientific
conversation.
Janna Levin (Professor of Physics and Astronomy)
01:50.560
And
so
I
want
to
begin
with
you,
Jan,
about
neural
nets.
Here's
this
instance
of
kind
of
biomimicry
where
you
have
these
computational
neural
networks
that
are
emulating
human
networks.
Can
you
describe
to
us
what
that
means
that
a
machine
is
emulating
human
neural
networks?
Yann LeCun (Chief AI Scientist)
02:09.640
Well,
it's
not
really
mimicry.
It's
more
inspiration
the
same
way
I
don't
know
airplanes
are
inspired
by
by
birds,
right?
Janna Levin (Professor of Physics and Astronomy)
02:19.200
And
the
That
didn't
work
I
thought.
Say
again.
But
I
thought
that
didn't
work
copying
birds
with
airplanes.
Yann LeCun (Chief AI Scientist)
02:25.240
Well,
in
the
sense
that
you
know
airplanes
have
wings
like
birds
and
they
generate
shift
by
propelling
themselves
through
the
air,
but
then
the
analogy
stops
stops
there.
And
the
wing
of
an
airplane
is
much
simpler
than
the
wing
of
a
bird,
but
yet
the
underlying
principle
is
the
Yann LeCun (Chief AI Scientist)
02:40.920
same.
Yann LeCun (Chief AI Scientist)
02:41.480
So,
neural
networks
are
a
bit
like
like
that,
like
are
like,
you
know,
our
two
real
brains
as
airplanes
are
two
birds.
They're
much
simplified
in
many
ways.
Um,
but
perhaps
some
of
the
underlying
principles
are
the
same.
We
don't
actually
know
because
we
don't
really
know
the
Yann LeCun (Chief AI Scientist)
03:00.400
sort
of
underlying
algorithm
of
the
cortex
if
you
want
or
the
method
by
which
the
brain
organizes
itself
and
learns.
So
Yann LeCun (Chief AI Scientist)
03:11.920
we
invented
substitutes.
Sort
of
like
birds
flap
their
wings
and
not
airplanes
right
the
air
propellers
so
or
turbojets.
In
our
nets
we
have
learning
algorithms
and
they
they
allow
artificial
nets
to
learn
in
a
way
that
we
think
is
similar
to
how
the
brains
learn.
Yann LeCun (Chief AI Scientist)
03:35.200
So,
the
brain
is
a
network
of
neurons,
the
neurons
are
interconnected
with
each
other,
and
the
way
the
brain
learns
is
by
modifying
the
efficacy
of
the
connections
between
the
neurons.
And
the
way
a
neural
net
is
trained
is
by
modifying
the
efficacy
of
the
connections
between
Yann LeCun (Chief AI Scientist)
03:51.000
those
similarity
neurons.
Yann LeCun (Chief AI Scientist)
03:52.720
Each
of
those
is
like
a
we
call
it
a
parameter.
You
you
see
this
in
the
price
the
number
of
parameters
of
a
neural
net,
right?
So
the
the
biggest
neural
net
at
the
moment
have,
you
know,
hundreds
of
billions
of
parameters,
if
not
more,
and
um
those
are
the
individual
Yann LeCun (Chief AI Scientist)
04:09.320
coefficients
that
are
modified
by
by
by
training.
So
Janna Levin (Professor of Physics and Astronomy)
04:14.440
And
how
is
deep
learning
uh
emerge
in
this
discussion?
Because
deep
learning
came
along
the
path
after
thinking
about
neural
nets.
And
this
has
been
since
the
80s
or
earlier
even.
Yann LeCun (Chief AI Scientist)
04:25.280
Um
yeah,
80s
roughly.
Um
So
early
neural
on
that,
the
first
ones
are
capable
of
learning
or
learning
something
useful
at
least
in
the
50s
were
shallow.
You
could
you
could
basically
train
a
single
layer
of
neurons,
right?
So
you
would
feed
the
input
and
train
the
system
to
Yann LeCun (Chief AI Scientist)
04:45.800
produce
a
particular
output
and
you
could
use
those
things
to
kind
of
recognize
or
classify
relatively
simple
patterns
but
not
really
sort
of
complex
things.
And
Yann LeCun (Chief AI Scientist)
04:57.000
people
at
the
time
even
in
the
60s
realise
that
the
way
to
make
progress
was
going
to
be
able
to
train
neural
nets
with
multiple
layers.
They
built
neural
nets
with
multiple
layers,
but
they
couldn't
train
all
the
layers,
so
we
only
trained
the
last
layer,
for
example.
And
Autoscroll