Toward Next-Generation Artificial Intelligence: Catalyzing the NeuroAI Revolution — part 1/3

Cleaned-text code points [0, 13327). Bibliography omitted here; source.md includes it.

Catalyzing Next-Generation Artificial Intelligence through NeuroAI
By Anthony Zador*, Sean Escola*, Blake Richards, Bence Ölveczky, Yoshua Bengio, Kwabena Boahen,
Matthew Botvinick,  Dmitri Chklovskii, Anne Churchland, Claudia Clopath, James DiCarlo, Surya
Ganguli, Jeff Hawkins, Konrad Körding, Alexei Koulakov, Yann LeCun, Timothy Lillicrap, Adam
Marblestone, Bruno Olshausen, Alexandre Pouget, Cristina Savin, Terrence Sejnowski, Eero Simoncelli,
Sara Solla, David Sussillo, Andreas S. Tolias, Doris Tsao
* Equal contributions
Abstract
:
Neuroscience
has
long
been
an
essential
driver
of
progress
in
artificial
intelligence
(AI).
We
propose
that
to
accelerate
progress
in
AI,
we
must
invest
in
fundamental
research
in
NeuroAI.
A
core
component
of
this
is
the
embodied
Turing
test,
which
challenges
AI
animal
models
to
interact
with
the
sensorimotor
world
at
skill
levels
akin
to
their
living
counterparts.
The
embodied
Turing
test
shifts
the
focus
from
those
capabilities
like
game
playing
and
language
that
are
especially
well-developed
or
uniquely
human
to
those
capabilities
–
inherited
from
over
500
million
years
of
evolution
–
that
are
shared
with
all
animals.
Building
models
that
can
pass
the
embodied
Turing
test
will
provide
a
roadmap
for the next generation of AI.
Over
the
coming
decades,
Artificial
Intelligence
(AI)
will
transform
society
and
the
world
economy
in
ways
that
are
as
profound
as
the
computer
revolution
of
the
last
half
century
and
likely
at
an
even
faster
pace.
This
AI
revolution
presents
tremendous
opportunities
to
unleash
human
creativity
and
catalyze
economic
growth,
relieving
workers
from
performing
the
most
dangerous
and
menial
jobs.
However,
to
reach
this
potential,
we
still
require
advances
that
will
make
AI
more
human-like
in
its
capabilities.
Historically,
neuroscience
has
been
a
critical
driver
and
source
of
inspiration
for
improvements
in
AI,
particularly
those
that
made
AI
more
proficient
in
areas
that
humans
and
other
animals
excel
at,
such
as
vision,
reward-based
learning,
interacting
with
the
physical
world,
and
language
1,2
.
It
can
still
play
this
role.
To
accelerate
progress
in
AI
and
realize
its
vast
potential,
we
must
invest
i
n
fundamental
research
in
“NeuroAI.”
The
seeds
of
the
current
AI
revolution
were
planted
decades
ago,
mainly
by
researchers
attempting
to
understand
how
brains
compute
3
.
Indeed,
the
earliest
efforts
to
build
an
“artificial
brain”
led
to
the
invention
of
the
modern
“von
Neumann
computer
architecture,”
for
which
John
von
Neumann
explicitly
drew
upon
the
very
limited
knowledge
of
the
brain
available
to
him
in
the
1940s
4,5
.
Later,
the
Nobel-prize
winning
work
of
David
Hubel
and
Torsten
Wiesel
on
visual
processing
circuits
in
the
cat
neocortex
inspired
the
deep
convolutional
networks
that
have
catalyzed
the
recent
revolution
in
modern
AI
6–8
.
Similarly,
the
development
of
reinforcement
learning
was
directly
inspired
by
insights
into
animal
behavior
and
neural
activity
during
learning
9–15
.
Now,
decades
later,
applications
of
ANNs
and
RL
are
coming
so
quickly
that
many
observers
assume
that
the
long-elusive
goal
of
human-level
intelligence
–
sometimes
referred
to
as
“artificial
general
intelligence”
–
is
within
our
grasp.
However,
in
contrast
to
the
optimism
of
those
outside
the
field,
many
front-line
AI
researchers
believe
that
major
breakthroughs
are

needed
before
we
can
build
artificial
systems
capable
of
doing
all
that
a
human,
or
even
a
much
simpler
animal like a mouse, can do.
Although
AI
systems
can
easily
defeat
any
human
opponent
in
games
such
as
chess
16
and
Go
17
,
they
are
not
robust
and
often
struggle
when
faced
with
novel
situations.
Moreover,
we
have
yet
to
build
effective
systems
that
can
walk
to
the
shelf,
take
down
the
chess
set,
set
up
the
pieces,
and
move
them
around
during
a
game,
although
recent
progress
is
encouraging
18
.
Similarly,
no
machine
can
build
a
nest,
forage
for
berries,
or
care
for
young.
Today’s
AI
systems
cannot
compete
with
the
sensorimotor
capabilities
of
a
four-year
old
child
or
even
simple
animals.
Many
basic
capacities
required
to
navigate
new
situations
–
capacities
that
animals
have
or
acquire
effortlessly
–
turn
out
to
be
deceptively
challenging
for
AI,
partly
because
AI
systems
lack
even
the
basic
abilities
to
interact
with
an
unpredictable
world.
A
growing
number
of
AI
researchers
doubt
that
merely
scaling
up
current
approaches
will
overcome
these
limitations.
Given
the
need
to
achieve
more
natural
intelligence
in
AI,
it
is
quite
likely
that
new
inspiration from naturally intelligent systems is needed
19
.
Historically,
many
key
AI
advances,
such
as
convolutional
ANNs
and
reinforcement
learning,
were
inspired
by
neuroscience.
Neuroscience
continues
to
provide
guidance
–
e.g.,
attention-based
neural
networks
were
loosely
inspired
by
attention
mechanisms
in
the
brain
20–23
–
but
this
is
often
based
on
findings
that
are
decades
old.
The
fact
that
such
cross-pollination
between
AI
and
neuroscience
is
far
less
common
than
in
the
past
represents
a
missed
opportunity.
Over
the
last
decades,
through
efforts
such
as
the
NIH
BRAIN
initiative
and
others,
we
have
amassed
an
enormous
amount
of
knowledge
about
the
brain.
The
emerging
field
of
NeuroAI,
at
the
intersection
of
neuroscience
and
AI,
is
based
on
the
premise
that
a
better
understanding
of
neural
computation
will
reveal
fundamental
ingredients
of
intelligence
and
catalyze
the
next
revolution
in
AI.
This
will
eventually
lead
to
artificial
agents
with
capabilities
that
match
those
of
humans.
The
NeuroAI
program
we
advocate
is
driven
by
the
recognition
that
AI
historically
owes
much
to
neuroscience
and
the
promise
that
AI
will
continue
to
learn
from
it–but
only
if
there
is
a
large
enough
community
of
researchers
fluent
in
both
domains.
We
believe
the
time
is
right
for
a
large-scale
effort
to
identify
and
understand
the
principles
of
biological
intelligence
and
abstract those for application in computer and robotic systems.
It
is
tempting
to
focus
on
the
most
characteristically
human
aspects
of
intelligent
behavior,
such
as
abstract
thought
and
reasoning.
However,
the
basic
ingredients
of
intelligence
–
adaptability,
flexibility,
and
the
ability
to
make
general
inferences
from
sparse
observations
–
are
already
present
in
some
form
in
basic
sensorimotor
circuits
which
have
been
evolving
for
hundreds
of
millions
of
years.
As
AI
pioneer
Hans
Moravec
24
put
it,
abstract
thought
“is
a
new
trick,
perhaps
less
than
100
thousand
years
old….effective
only
because
it
is
supported
by
this
much
older
and
much
more
powerful,
though
usually
unconscious,
sensorimotor
knowledge.”
This
implies
that
the
bulk
of
the
work
in
developing
general
AI
can
be
achieved
by
building
systems
that
match
the
perceptual
and
motor
abilities
of
animals
and
that
the
subsequent
step
to
human-level
intelligence
would
be
considerably
smaller.
This
is
good
news
because
progress
on
the
first
goal
can
rely
on
the
favored
subjects
of
neuroscience
research
–
rats,
mice,
and
non-human
primates
–
for
which
extensive
and
rapidly
expanding
behavioral
and
neural
datasets
can
guide
the
way.
Thus,
we
believe
that
the
NeuroAI
path
will
lead
to
necessary
advances
if
we
figure
out
the
core capabilities that all animals possess in embodied sensorimotor interaction with the world.

NeuroAI Grand Challenge: The Embodied Turing Test
In
1950,
Alan
Turing
proposed
the
“imitation
game”
as
a
test
of
a
machine’s
ability
to
exhibit
intelligent
behavior
indistinguishable
from
that
of
a
human
25
(Figure
1,
left).
In
that
game,
now
known
as
the
Turing
test,
a
human
judge
evaluates
natural
language
conversations
between
a
real
human
and
a
machine
trained
to
mimic
human
responses.
By
focusing
on
conversational
abilities,
Turing
evaded
asking
whether
a
machine
could
“think,”
a
question
he
considered
impossible
to
answer.
The
Turing
test
is
based
on
the
implicit
belief
that
language
represents
the
pinnacle
of
human
intelligence
and
that
a
machine
capable
of
conversation must surely be intelligent.
Until
recently,
no
artificial
system
could
come
close
to
passing
the
Turing
test.
However,
a
class
of
modern
AI
systems
called
“large
language
models”
can
now
engage
in
surprisingly
cogent
conversations
26
.
In
part,
their
success
reveals
how
easily
we
can
be
tricked
into
imputing
intelligence,
agency,
and
even
consciousness
to
our
interlocutor
27
.
Impressive
though
these
systems
are,
because
they
are
not
grounded
in
real-world
experiences,
they
nonetheless
continue
to
struggle
with
many
basic
aspects
of
causal
reasoning
and
physical
common
sense.
Thus,
the
Turing
test
does
not
probe
our
prodigious
perceptual
and
motor
abilities
to
interact
with
and
reason
about
the
physical
world,
abilities
shared
with
animals and honed through countless generations of natural selection.
We
therefore
propose
an
expanded
“embodied
Turing
test,”
one
that
includes
advanced
sensorimotor
abilities
(Figure
1,
right).
The
spirit
of
the
original
Turing
test
was
to
establish
a
simple
qualitative
standard
against
which
our
progress
toward
building
artificially
intelligent
machines
can
be
judged.
This
embodied
Turing
test
would
benchmark
and
compare
the
interactions
with
the
world
of
artificial
systems
versus
humans
and
other
animals.
Similar
ideas
have
been
proposed
previously
28–32
.
However,
in
light
of
recent
advances
enabling
large-scale
behavioral
and
neural
measurements,
as
well
as
large-scale
simulations
of
embodied
agents
in
silico,
we
believe
the
time
is
ripe
to
instantiate
a
major
research
effort
in
this
direction.
Because
each
animal
has
its
own
unique
set
of
abilities,
each
animal
defines
its
own
embodied
Turing
test:
An
artificial
beaver
might
be
tested
on
its
ability
to
build
a
dam,
and
an
artificial
squirrel
on
its
ability
to
jump
through
trees.
Nonetheless,
many
core
sensorimotor
capabilities
are
shared
by
almost
all
animals,
and
the
ability
of
animals
to
rapidly
evolve
the
sensorimotor
skills
needed
to
adapt
to
new
environments
suggests
that
these
core
skills
provide
a
solid
foundation.
This
implies
that
after
developing
an
AI
system
to
faithfully
reproduce
the
behavior
of
one
species,
the
adaptation
of
this
system
to
other
species
–
and
even
to
humans
–
may
be
straightforward.
Below
we
highlight
a
few
of
the
characteristics that are shared across species.

Figure
1.
Turing
tests:
comparisons
between
the
performance
of
AI
systems
and
their
living
counterparts.
Left
:
The
original
Turing
test
as
proposed
by
Alan
Turing
25
.
If
a
human
tester
cannot
determine
whether
their
interlocutor
is
an
AI
system
or
another
human,
the
AI
passes
the
test.
Modern
large
language
models
have
made
substantial
progress
towards
passing
this
test
26
.
Right
:
The
embodied
Turing
test.
An
AI
animal
model
–
whether
robotic
or
in
simulation
–
passes
the
test
if
its
behavior
is
indistinguishable
from
that
of
its
living
counterpart.
No
AI
systems
are
close
to
passing
this
test.
Here,
an
artificial
beaver
is
tested on the species-specific behavior of dam construction.
Animals
engage
their
environments.
The
defining
feature
of
animals
is
their
ability
to
move
around
and
interact
with
their
environment
in
purposeful
ways.
Despite
recent
advances
in
optimal
control,
reinforcement
learning,
and
imitation
learning,
robotics
is
still
far
from
achieving
animal-level
abilities
in
controlling
their
bodies
and
manipulating
objects,
even
in
simulation.
Of
course,
neuroscience
can
provide
guidance
about
the
kinds
of
modular
and
hierarchical
architectures
that
could
be
adapted
to
artificial
systems
to
give
them
these
capabilities
33
.
It
can
also
provide
us
with
design
principles
like
partial
autonomy
(how
low-level
modules
in
a
hierarchy
act
semi-autonomously
in
the
absence
of
input
from
high-level
modules)
and
amortized
control
(how
movements
generated
at
first
by
a
slow
planning
process
are
eventually
transferred
to
a
fast
reflexive
system).
These
principles
could
guide
the
design
of
systems
for
perception,
action
selection,
locomotion,
and
fine-grained
control
of
limbs,
hands,
and
fingers.
Understanding
how
specific
neural
circuits
participate
in
different
tasks
could
also
inspire
solutions
for
other
forms
of
‘intelligence,’
including
in
more
cognitive
realms.
For
example,
we
speculate
that
incorporating
principles
of
circuitry
for
low-level
motor
control
could
help
provide
a
better
basis
for
higher-level motor planning in AI systems.
Animals
behave
flexibly.
Another
goal
is
to
develop
AI
systems
that
can
engage
a
large
repertoire
of
flexible
and
diverse
tasks
in
a
manner
that
echoes
the
incredible
range
of
behaviors
that
individual
animals
can
generate.
Modern
AI
can
easily
learn
to
outperform
humans
at
video
games
like
Breakout
using
nothing
more
than
pixels
on
a
screen
and
game
scores
34
.
However,
these
systems,
unlike
human
players,
are
brittle
and
highly
sensitive
to
small
perturbations:
changing
the
rules
of
the
game
slightly,
or
even
a
few
pixels
on
the
input,
can
lead
to
catastrophically
poor
performance
35
.
This
is
because
these
systems
learn
a
mapping
from
pixels
to
actions
that
need
not
involve
an
understanding
of
the
agents
and
objects
in
the
game
and
the
physics
that
governs
them.
Similarly,
a
self-driving
car
does
not
inherently
know
about
the
danger
of
a
crate
falling
off
a
truck
in
front
of
it
unless
it
has
literally
seen
examples
of