TRANSCRIPT 🔗
WEBVTT
00:00:00.400 --> 00:00:09.210
Welcome to Sleepy Facts about the Universe. We
00:00:12.890 --> 00:00:15.710
often carry an unexamined picture of intelligence
00:00:15.710 --> 00:00:19.350
as something sealed inside the skull, a private
00:00:19.350 --> 00:00:22.760
possession carried around by the body. But what
00:00:22.760 --> 00:00:25.940
if this familiar model is inaccurate, and intelligence
00:00:25.940 --> 00:00:28.820
is better understood as an activity that happens
00:00:28.820 --> 00:00:32.140
rather than a thing that is held? This episode
00:00:32.140 --> 00:00:35.829
explores the process view of cognition. examining
00:00:35.829 --> 00:00:38.869
how intelligent behavior emerges from the sustained
00:00:38.869 --> 00:00:42.630
coupling of brain, body, tools, and environment.
00:00:43.090 --> 00:00:47.250
To explain this shift in perspective, the episode
00:00:47.250 --> 00:00:51.170
looks at several revealing examples. We consider
00:00:51.170 --> 00:00:54.409
how an orb -weaving spider's web functions as
00:00:54.409 --> 00:00:57.509
an extension of her sensory system, how solving
00:00:57.509 --> 00:01:00.509
complex multiplication on paper relies on external
00:01:00.509 --> 00:01:03.729
symbols to do cognitive work, and how an ant
00:01:03.729 --> 00:01:06.590
colony processes information without a central
00:01:06.590 --> 00:01:09.989
controller. We also examine the extended mind
00:01:09.989 --> 00:01:12.969
thesis proposed by Andy Clark and David Chalmers,
00:01:13.109 --> 00:01:16.170
which argues that reliably integrated tools can
00:01:16.170 --> 00:01:19.170
literally function as part of our memory, alongside
00:01:19.170 --> 00:01:22.030
the practical and philosophical objections raised
00:01:22.030 --> 00:01:25.890
by critics of this view. I kept returning to
00:01:25.890 --> 00:01:28.890
one question while working on this episode. Why
00:01:28.890 --> 00:01:31.290
do we assume the boundary of the mind stops at
00:01:31.290 --> 00:01:34.620
the skin? Looking at the dynamic relationships
00:01:34.620 --> 00:01:38.040
between organisms and their surroundings offers
00:01:38.040 --> 00:01:41.060
a messier but potentially more faithful account
00:01:41.060 --> 00:01:45.560
of how thinking actually works. It is a proposal
00:01:45.560 --> 00:01:48.159
that changes what we look for when we try to
00:01:48.159 --> 00:01:51.959
understand intelligence, inviting us to consider
00:01:51.959 --> 00:01:55.859
that much of what we do, we do with and through
00:01:55.859 --> 00:02:01.700
the world. We carry a picture we never chose,
00:02:02.019 --> 00:02:04.659
intelligence as something a person possesses,
00:02:04.969 --> 00:02:08.250
like height or eye color, sealed inside the skull
00:02:08.250 --> 00:02:10.830
and carried about by the body like a vehicle.
00:02:11.310 --> 00:02:14.349
But what if the familiar picture is not accurate?
00:02:15.099 --> 00:02:18.300
What if intelligence is better understood not
00:02:18.300 --> 00:02:22.300
as a possession at all, but as an activity, something
00:02:22.300 --> 00:02:25.099
that happens only in the sustained coupling of
00:02:25.099 --> 00:02:29.680
brain, body, tool and world? There is a picture
00:02:29.680 --> 00:02:32.580
most of us carry without ever having chosen it.
00:02:33.129 --> 00:02:36.169
Intelligence is something a person has, like
00:02:36.169 --> 00:02:39.689
height or eye color. It sits inside the skull,
00:02:39.990 --> 00:02:44.090
private and contained, and the body is merely
00:02:44.090 --> 00:02:47.490
the vehicle that carries it around. The world
00:02:47.490 --> 00:02:52.189
outside is scenery, input, distraction. This
00:02:52.189 --> 00:02:54.430
picture feels obvious because it is familiar,
00:02:54.750 --> 00:02:58.159
but familiarity is not the same as accuracy.
00:02:58.849 --> 00:03:01.710
And tonight I want to ask whether intelligence
00:03:01.710 --> 00:03:04.550
might be better understood. Not as a possession,
00:03:04.689 --> 00:03:08.530
but as an activity. Something that happens rather
00:03:08.530 --> 00:03:12.400
than something that is hell. The question matters
00:03:12.400 --> 00:03:14.819
because it changes what we look for when we try
00:03:14.819 --> 00:03:18.349
to explain how thinking works. If intelligence
00:03:18.349 --> 00:03:21.110
is a thing, then the explanation lies inside
00:03:21.110 --> 00:03:24.590
the object that possesses it. We open the skull,
00:03:24.830 --> 00:03:29.620
examine the brain, measure its parts. If intelligence
00:03:29.620 --> 00:03:33.219
is a process, then the explanation lies in relationships.
00:03:34.379 --> 00:03:37.930
Between an agent and its body. Between the body
00:03:37.930 --> 00:03:41.949
and tools. Between tools and an environment.
00:03:42.840 --> 00:03:46.830
Between one agent and another. With feedback
00:03:46.830 --> 00:03:49.330
returning at each stage to alter what happens
00:03:49.330 --> 00:03:54.620
next. The first picture is tidy. The second is
00:03:54.620 --> 00:03:57.849
messier. But it may be more faithful to what
00:03:57.849 --> 00:04:01.780
we actually observe. Let me propose a working
00:04:01.780 --> 00:04:05.840
description. Intelligence, understood as a process,
00:04:06.120 --> 00:04:09.639
is sensing, comparing, remembering, acting, receiving
00:04:09.639 --> 00:04:13.270
feedback and revising behavior. It is sustained
00:04:13.270 --> 00:04:16.920
across time, not a single event. It resembles
00:04:16.920 --> 00:04:19.779
metabolism or flight more than it resembles a
00:04:19.779 --> 00:04:24.000
jewel in a box. Metabolism is not an ingredient
00:04:24.000 --> 00:04:27.300
you can extract and hold. It is the organized
00:04:27.300 --> 00:04:30.240
activity by which a living system maintains itself.
00:04:30.730 --> 00:04:34.670
Flight is not located in the wing alone. It is
00:04:34.670 --> 00:04:37.990
the dynamic relationship between wing, air, muscle,
00:04:38.269 --> 00:04:40.990
nervous system and
Load full transcript
the continuous adjustments
00:04:40.990 --> 00:04:45.629
that keep a body aloft. We do not ask where flight
00:04:45.629 --> 00:04:50.649
is stored. We describe how it is achieved. The
00:04:50.649 --> 00:04:53.129
proposal before us is that intelligence deserves
00:04:53.129 --> 00:04:57.040
the same treatment. Consider the orb -weaving
00:04:57.040 --> 00:05:00.759
spider. The spider sits at the center of a web
00:05:00.759 --> 00:05:04.629
or near it and waits. When an insect strikes
00:05:04.629 --> 00:05:08.509
the web, the spider does not see the insect the
00:05:08.800 --> 00:05:12.959
web does the seeing for her vibrations travel
00:05:12.959 --> 00:05:17.180
through silk threads the spider reads location
00:05:17.180 --> 00:05:20.759
size and perhaps the nature of the prey from
00:05:20.759 --> 00:05:24.319
the pattern of those vibrations the web is not
00:05:24.319 --> 00:05:28.230
merely a trap it is an extension of the spider's
00:05:28.230 --> 00:05:32.120
sensory system into the environment the density
00:05:32.120 --> 00:05:36.100
and tension of the silk the geometry of the radial
00:05:36.100 --> 00:05:39.699
and spiral threads, the way vibrations dampen
00:05:39.699 --> 00:05:43.199
or propagate. All of this is part of the information
00:05:43.199 --> 00:05:46.319
processing by which the spider discriminates
00:05:46.319 --> 00:05:50.740
food from debris, mate from threat. Researchers
00:05:50.740 --> 00:05:53.579
who study this have argued, carefully and with
00:05:53.579 --> 00:05:56.459
evidence, that the web functions as part of the
00:05:56.459 --> 00:06:00.230
spider's cognitive system, not as a tool the
00:06:00.230 --> 00:06:03.269
spider uses the way a carpenter uses a hammer.
00:06:03.980 --> 00:06:07.970
something more integrated. The spider's body
00:06:07.970 --> 00:06:12.410
ends at her legs, but her sensing does not. Whether
00:06:12.410 --> 00:06:15.389
her cognition also extends into the web is a
00:06:15.389 --> 00:06:19.160
proposal, not a settled fact. But the boundary
00:06:19.160 --> 00:06:23.420
is no longer obvious. It is instead a question.
00:06:24.180 --> 00:06:28.540
Now shift scale. A person sits with pencil and
00:06:28.540 --> 00:06:32.060
paper, solving a multiplication problem too large
00:06:32.060 --> 00:06:36.560
to hold in working memory. Perhaps 347 times,
00:06:36.819 --> 00:06:41.550
56, the unaided brain can manage some arithmetic.
00:06:41.689 --> 00:06:46.930
But not this. Not reliably. Not without external
00:06:46.930 --> 00:06:50.769
support. So the person writes partial products,
00:06:51.069 --> 00:06:54.189
carries digits, arranges symbols in columns.
00:06:54.490 --> 00:06:57.930
The paper holds memory that the brain cannot.
00:06:58.350 --> 00:07:01.839
The pencil executes transformations that the
00:07:01.839 --> 00:07:05.800
unaided mind would struggle to track. The numerals
00:07:05.800 --> 00:07:09.029
are not mere prompts. They are doing cognitive
00:07:09.029 --> 00:07:14.370
work. The person, the pencil, the paper, the
00:07:14.370 --> 00:07:17.600
spatial arrangement of symbols, the culturally
00:07:17.600 --> 00:07:21.000
learned algorithm. All of these together constitute
00:07:21.000 --> 00:07:25.050
the calculating system. Remove any part and the
00:07:25.050 --> 00:07:27.850
multiplication fails or becomes a different,
00:07:28.050 --> 00:07:32.980
harder task. In 1998, philosophers Andy Clark
00:07:32.980 --> 00:07:35.500
and David Chalmers published a paper that made
00:07:35.500 --> 00:07:39.170
this case explicit. They imagined two people,
00:07:39.910 --> 00:07:43.009
Inger, who remembers the location of a museum
00:07:43.009 --> 00:07:46.980
by consulting her biological memory, and Otto,
00:07:47.220 --> 00:07:50.259
who has memory impairment and uses a notebook
00:07:50.259 --> 00:07:52.699
to record and retrieve the same information.
00:07:53.600 --> 00:07:56.879
The notebook, Clark and Chalmers argued, plays
00:07:56.879 --> 00:07:59.519
the same functional role for Otto that biological
00:07:59.519 --> 00:08:03.120
memory plays for Inger. If we are willing to
00:08:03.120 --> 00:08:05.920
say Inga knows where the museum is, they asked,
00:08:06.060 --> 00:08:09.139
why should we deny that Otto knows, merely because
00:08:09.139 --> 00:08:13.350
his memory is external? The proposal is not that
00:08:13.350 --> 00:08:16.189
any object a person touches becomes part of their
00:08:16.189 --> 00:08:19.240
mind. The notebook must be reliably available,
00:08:19.660 --> 00:08:22.720
automatically consulted, and trusted in roughly
00:08:22.720 --> 00:08:26.600
the way biological memory is trusted, functionally
00:08:26.600 --> 00:08:29.800
integrated into Otto's cognitive routines, not
00:08:29.800 --> 00:08:33.029
merely useful. But when those conditions are
00:08:33.029 --> 00:08:36.009
met, the extended mind thesis suggests that the
00:08:36.009 --> 00:08:39.409
boundary of cognition moves outward. The proposal
00:08:39.409 --> 00:08:43.570
remains contested. Critics ask whether a notebook
00:08:43.570 --> 00:08:47.110
that can be lost, stolen, or left at home is
00:08:47.110 --> 00:08:50.149
truly equivalent to neural tissue that cannot
00:08:50.149 --> 00:08:53.919
be separated from the person. The debate continues.
00:08:54.870 --> 00:08:58.490
What is clear, even without settling the metaphysics,
00:08:58.940 --> 00:09:01.360
Is that much of what we call intelligent behavior
00:09:01.360 --> 00:09:04.480
in humans depends on this kind of coupling between
00:09:04.480 --> 00:09:09.110
internal resources and external structures? Consider
00:09:09.110 --> 00:09:12.690
next a robot. Perhaps one designed to navigate
00:09:12.690 --> 00:09:15.549
a warehouse, identify objects, and move them
00:09:15.549 --> 00:09:19.370
to designated locations. The robot appears competent.
00:09:19.970 --> 00:09:24.450
It avoids obstacles, adjusts to variations, completes
00:09:24.450 --> 00:09:28.679
its tasks.
but this competence is not contained
00:09:28.679 --> 00:09:31.799
in its processor the way a program might be contained
00:09:31.799 --> 00:09:35.269
on a disk. The robot's cameras assumed certain
00:09:35.269 --> 00:09:38.860
lighting conditions. Its training data assumed
00:09:38.860 --> 00:09:42.000
certain floor surfaces, certain object shapes,
00:09:42.320 --> 00:09:46.070
certain distributions of clutter. Its software
00:09:46.070 --> 00:09:48.309
was written with assumptions about the physics
00:09:48.309 --> 00:09:52.159
of wheels and the friction of concrete. Change
00:09:52.159 --> 00:09:54.639
the floor from concrete to polished tile and
00:09:54.639 --> 00:09:58.539
the robot may freeze or fall. Not because its
00:09:58.539 --> 00:10:01.460
processor has failed, but because the coupled
00:10:01.460 --> 00:10:04.220
system of machine and environment has been disrupted.
00:10:04.659 --> 00:10:09.649
A moved shelf. A lighting failure. These are
00:10:09.649 --> 00:10:13.539
not mere inconveniences. They are breakdowns
00:10:13.539 --> 00:10:15.779
of the relationship that sustained the robot's
00:10:15.779 --> 00:10:19.240
apparent competence. The intelligence was never
00:10:19.240 --> 00:10:23.169
only inside the robot. It was always a process
00:10:23.169 --> 00:10:26.230
distributed across the machine and the environment
00:10:26.230 --> 00:10:30.659
its designers anticipated. Now consider the ant
00:10:30.659 --> 00:10:34.919
colony. No single ant knows the location of all
00:10:34.919 --> 00:10:38.639
food sources, the state of all chambers, the
00:10:38.639 --> 00:10:42.820
needs of all larvae. Each ant follows local rules,
00:10:43.100 --> 00:10:46.659
responds to chemical signals, touches antanny
00:10:46.659 --> 00:10:50.529
with others, lays trails, changes behavior based
00:10:50.529 --> 00:10:54.720
on brief encounters. From these local interactions,
00:10:54.919 --> 00:10:59.090
colony -level behavior emerges. Foraging routes
00:10:59.090 --> 00:11:04.210
are established. Nest sites are selected. Labor
00:11:04.210 --> 00:11:07.990
is divided. The colony adapts to disturbance.
00:11:09.309 --> 00:11:12.549
Biologist Deborah Gordon, who has studied harvester
00:11:12.549 --> 00:11:15.549
ants for decades, emphasizes that there is no
00:11:15.549 --> 00:11:18.110
central representation of the colony's state.
00:11:19.049 --> 00:11:23.480
No ant who decides. The colony's behavior is
00:11:23.480 --> 00:11:25.879
distributed across the interactions of hundreds
00:11:25.879 --> 00:11:28.779
or thousands of individuals, each with limited
00:11:28.779 --> 00:11:31.659
information, each altering the environment that
00:11:31.659 --> 00:11:35.909
the next ant encounters. This does not mean the
00:11:35.909 --> 00:11:39.049
colony has a unified mind or conscious experience.
00:11:39.669 --> 00:11:42.950
It means that adaptive information processing
00:11:42.950 --> 00:11:46.190
behavior can emerge from distributed interactions
00:11:46.190 --> 00:11:50.610
without a central controller. Whether this counts
00:11:50.610 --> 00:11:53.149
as cognition depends on where we draw the boundary.
00:11:53.679 --> 00:11:57.440
If cognition requires a single brain, then the
00:11:57.440 --> 00:12:00.919
colony is not cognitive. If cognition is the
00:12:00.919 --> 00:12:03.500
organized processing of information that produces
00:12:03.500 --> 00:12:06.860
adaptive behavior, then the colony begins to
00:12:06.860 --> 00:12:09.659
look like a cognitive system spread across many
00:12:09.659 --> 00:12:13.610
bodies and their chemical traces. These examples
00:12:13.610 --> 00:12:16.649
operate at different scales, but they share a
00:12:16.649 --> 00:12:19.889
common feature. In each case, the intelligent
00:12:19.889 --> 00:12:22.850
behavior we observe is not located in a single
00:12:22.850 --> 00:12:26.840
component. It emerges from relationships, feedback
00:12:26.840 --> 00:12:29.799
loops, and sustained couplings between parts.
00:12:30.639 --> 00:12:35.360
The spider and her web. The person and her symbols.
00:12:36.090 --> 00:12:40.490
The robot and its assumed environment. The colony
00:12:40.490 --> 00:12:44.629
and its network of encounters. The question is
00:12:44.629 --> 00:12:47.289
not whether brains, processors, or individual
00:12:47.289 --> 00:12:51.649
ants matter. They do. The question is whether
00:12:51.649 --> 00:12:55.090
they matter in isolation, or whether they matter
00:12:55.090 --> 00:12:57.769
as parts of larger processes that extend beyond
00:12:57.769 --> 00:13:01.330
their physical boundaries. This brings us to
00:13:01.330 --> 00:13:03.830
two related proposals that have emerged from
00:13:03.830 --> 00:13:06.429
cognitive science and philosophy over the past
00:13:06.429 --> 00:13:10.250
several decades. The first is embodied cognition.
00:13:10.889 --> 00:13:14.450
This is the claim that cognitive processes depend
00:13:14.450 --> 00:13:18.029
constitutively on the body's structure, movement,
00:13:18.580 --> 00:13:22.179
and interaction with the world not that the body
00:13:22.179 --> 00:13:24.799
merely carries the brain around but that the
00:13:24.799 --> 00:13:28.460
shape of our hands the layout of our senses our
00:13:28.460 --> 00:13:32.159
capacity for physical action all enter into how
00:13:32.159 --> 00:13:36.679
we think the second is the extended mind this
00:13:36.740 --> 00:13:38.919
is the stronger claim that cognitive processes
00:13:38.919 --> 00:13:41.700
can literally extend beyond the body into tools
00:13:41.700 --> 00:13:44.879
notations and environmental structures these
00:13:45.299 --> 00:13:49.149
are proposals not settled verdicts They have
00:13:49.149 --> 00:13:51.610
defenders and critics and the debate is ongoing.
00:13:52.110 --> 00:13:56.500
The critics raise an important objection. There
00:13:56.500 --> 00:13:59.639
must be some boundary, they argue, between the
00:13:59.639 --> 00:14:04.120
cognitive system
and the tools it uses. A calculator
00:14:04.120 --> 00:14:06.899
assists cognition, but is it part of the mind?
00:14:07.549 --> 00:14:11.129
A map helps navigation, but is it part of the
00:14:11.129 --> 00:14:15.240
memory system? If we extend cognition too far,
00:14:15.580 --> 00:14:18.240
We risk losing the distinction between thinking
00:14:18.240 --> 00:14:20.500
and everything that thinking interacts with.
00:14:20.820 --> 00:14:23.879
This is the boundary problem, and it is genuine.
00:14:24.750 --> 00:14:28.809
The critics are not merely conservative. They
00:14:28.809 --> 00:14:31.570
point out that not every causal influence on
00:14:31.570 --> 00:14:34.950
thought is part of thought. The oxygen in the
00:14:34.950 --> 00:14:39.409
room enables cognition, but we do not call oxygen
00:14:39.409 --> 00:14:42.960
part of the mind. The question is where to draw
00:14:42.960 --> 00:14:47.059
the line and on what basis. The defenders of
00:14:47.059 --> 00:14:49.840
extended and embodied cognition have a response,
00:14:50.220 --> 00:14:53.059
and it is practical rather than metaphysical.
00:14:53.299 --> 00:14:57.149
The boundary we draw should depend on what we
00:14:57.149 --> 00:15:01.129
want to explain and predict. If we want to understand
00:15:01.129 --> 00:15:04.850
how the spider catches prey, we gain explanatory
00:15:04.850 --> 00:15:07.490
power by treating the web as part of the sensory
00:15:07.490 --> 00:15:11.389
system. If we want to understand how humans solve
00:15:11.389 --> 00:15:14.509
complex problems, we gain by attending to the
00:15:14.509 --> 00:15:18.200
external symbols and tools they recruit if we
00:15:18.200 --> 00:15:21.440
want to understand robot behavior we must include
00:15:21.440 --> 00:15:23.840
the training environment and its assumptions
00:15:23.840 --> 00:15:28.669
the boundary is not arbitrary but it is not fixed
00:15:28.669 --> 00:15:32.129
by nature either it is a methodological choice
00:15:32.129 --> 00:15:35.470
guided by which carving of the world yields the
00:15:35.470 --> 00:15:38.110
clearest explanations and the most reliable predictions
00:15:39.600 --> 00:15:42.100
The process view does not claim that everything
00:15:42.100 --> 00:15:45.639
is connected to everything else. It claims that
00:15:45.639 --> 00:15:47.639
the most useful boundaries for understanding
00:15:47.639 --> 00:15:51.899
intelligence often fall outside the skull. This
00:15:51.899 --> 00:15:54.919
reframing has consequences for how we think about
00:15:54.919 --> 00:15:59.320
several domains. In animal cognition, it directs
00:15:59.320 --> 00:16:02.059
attention to the structures animals build and
00:16:02.059 --> 00:16:05.970
the ways they modify their environments. Beavers,
00:16:06.049 --> 00:16:09.389
termites. Bowerbirds and many others are not
00:16:09.389 --> 00:16:13.250
merely behaving in an environment. They are constructing
00:16:13.250 --> 00:16:15.570
the environmental conditions under which their
00:16:15.570 --> 00:16:19.570
behavior makes sense. In human development, it
00:16:19.570 --> 00:16:22.149
suggests that learning to think is partly learning
00:16:22.149 --> 00:16:26.610
to arrange and use external resources. Language,
00:16:26.870 --> 00:16:30.070
writing, number systems, instruments, institutions.
00:16:31.659 --> 00:16:34.519
In artificial intelligence, it cautions against
00:16:34.519 --> 00:16:37.039
evaluating a system's intelligence in abstraction
00:16:37.039 --> 00:16:39.639
from the environments and assumptions that sustain
00:16:39.639 --> 00:16:43.029
it. A system that appears intelligent in one
00:16:43.029 --> 00:16:46.399
setting may fail in another, not because it has
00:16:46.399 --> 00:16:49.299
lost some internal capacity, but because the
00:16:49.299 --> 00:16:53.100
coupled process has been disrupted. The process
00:16:53.100 --> 00:16:56.519
view also preserves something important, the
00:16:56.620 --> 00:17:00.000
recognition that intelligence is hard -won, sustained
00:17:00.000 --> 00:17:03.570
and vulnerable. It does not emerge automatically
00:17:03.570 --> 00:17:07.390
from having the right parts. It must be maintained
00:17:07.390 --> 00:17:10.289
through continuous interaction, feedback, and
00:17:10.289 --> 00:17:14.650
adjustment. A spider without a web is not a spider
00:17:14.650 --> 00:17:18.369
with sensing temporarily disabled. She is a different
00:17:18.369 --> 00:17:21.329
kind of entity in a different kind of relation
00:17:21.329 --> 00:17:25.759
to her world. A person without paper and pencil,
00:17:25.980 --> 00:17:29.119
or without language, or without the cultural
00:17:29.119 --> 00:17:32.660
inheritance of mathematics can still think. but
00:17:32.660 --> 00:17:36.390
not in the same ways the process is not infinitely
00:17:36.390 --> 00:17:40.599
flexible it depends on stable structures reliable
00:17:40.599 --> 00:17:44.000
couplings and the capacity to revise when feedback
00:17:44.000 --> 00:17:48.069
demands it we should be careful not to overstate
00:17:48.069 --> 00:17:51.779
what has been established the extended mind thesis
00:17:51.779 --> 00:17:54.339
is a philosophical proposal not an empirical
00:17:54.339 --> 00:17:58.440
finding confirmed by neuroscience embodied cognition
00:17:59.160 --> 00:18:02.039
is a broad research program with multiple, sometimes
00:18:02.039 --> 00:18:06.180
conflicting versions. The spider web example
00:18:06.180 --> 00:18:09.559
shows that sensing can be externalized. It does
00:18:09.559 --> 00:18:13.130
not prove that all cognition is extended. The
00:18:13.130 --> 00:18:15.829
ant colony example shows distributed behavior.
00:18:16.319 --> 00:18:19.500
It does not prove that the colony has a unified
00:18:19.500 --> 00:18:23.420
mind or conscious experience. The robot example
00:18:23.420 --> 00:18:27.789
shows environmental dependence. It does not establish
00:18:27.789 --> 00:18:31.009
that robots are intelligent in any sense
comparable
00:18:31.009 --> 00:18:35.480
to human intelligence. What these examples do,
00:18:35.619 --> 00:18:39.220
collectively, is shift the burden of proof. They
00:18:39.220 --> 00:18:41.859
make it harder to assume without argument that
00:18:41.859 --> 00:18:44.460
intelligence is simply a property of isolated
00:18:44.460 --> 00:18:48.309
objects. And they invite us to consider whether
00:18:48.309 --> 00:18:51.150
the process view offers better explanations of
00:18:51.150 --> 00:18:54.910
the behaviors we actually observe. Let me return
00:18:54.910 --> 00:18:58.569
to the picture we began with. The mind is a private
00:18:58.569 --> 00:19:01.690
possession in a skull, looking out at the world
00:19:01.690 --> 00:19:05.009
through the senses, issuing commands to a passive
00:19:05.009 --> 00:19:09.839
body. This picture has a history. It is associated
00:19:09.839 --> 00:19:12.559
with early modern philosophy, particularly with
00:19:12.559 --> 00:19:15.099
Rene Descartes and his distinction between thinking
00:19:15.099 --> 00:19:18.859
substance and extended substance, between the
00:19:18.859 --> 00:19:21.539
mind that thinks and the body that occupies space.
00:19:22.799 --> 00:19:25.660
The picture was reinforced by the cognitive science
00:19:25.660 --> 00:19:28.680
of the mid -20th century, which often treated
00:19:28.680 --> 00:19:31.160
the brain as an information processing device,
00:19:31.500 --> 00:19:35.180
a computer made of neural tissue with the body
00:19:35.180 --> 00:19:38.950
as peripheral input -output equipment. These
00:19:38.950 --> 00:19:42.599
ideas have produced genuine insights, but they
00:19:42.599 --> 00:19:46.440
may also have obscured something important. That
00:19:46.440 --> 00:19:49.420
thinking is something we do, not merely something
00:19:49.420 --> 00:19:53.690
we have. And that much of what we do We do with
00:19:53.690 --> 00:19:57.660
and through the world. The idea that thinking
00:19:57.660 --> 00:20:00.200
is something we do with and through the world
00:20:00.200 --> 00:20:03.740
has older roots. But tonight's episode is not
00:20:03.740 --> 00:20:07.059
their history. It is an invitation to consider
00:20:07.059 --> 00:20:09.880
whether the process view, however arrived at,
00:20:10.019 --> 00:20:12.720
better captures what intelligence actually is.
00:20:13.660 --> 00:20:16.920
I want to close with an image that tests the
00:20:16.920 --> 00:20:20.490
mental model we have been constructing. Imagine
00:20:20.490 --> 00:20:24.299
an alien intelligence. Not the familiar science
00:20:24.299 --> 00:20:27.880
fiction figure of a single advanced being, but
00:20:27.880 --> 00:20:31.589
something distributed. Mobile organisms move
00:20:31.589 --> 00:20:34.630
through a habitat, encountering structures that
00:20:34.630 --> 00:20:37.549
store information, leaving traces that alter
00:20:37.549 --> 00:20:41.359
the behavior of the next organisms to pass. The
00:20:41.359 --> 00:20:43.880
habitat itself is modified by the organisms,
00:20:44.180 --> 00:20:47.259
and the modifications store memory of past encounters,
00:20:47.720 --> 00:20:53.190
an archive perhaps crystalline or chemical. accumulates
00:20:53.190 --> 00:20:56.390
records that no single organism consults in full,
00:20:56.609 --> 00:20:59.309
but that shape the conditions under which all
00:20:59.309 --> 00:21:03.029
organisms act. The intelligence of this system
00:21:03.029 --> 00:21:06.710
is impossible to locate in any component. It
00:21:06.710 --> 00:21:08.950
exists in the sustained process of interaction,
00:21:09.369 --> 00:21:12.430
feedback, and revision across the whole coupled
00:21:12.430 --> 00:21:16.029
system. It is not conscious in any way we could
00:21:16.029 --> 00:21:19.660
verify. It may not be unified in the way a single
00:21:19.660 --> 00:21:23.779
mind seems unified to its owner. but it is organized,
00:21:24.140 --> 00:21:28.339
adaptive, and sustained across time. It is, on
00:21:28.339 --> 00:21:32.279
the process view, intelligent. This image is
00:21:32.279 --> 00:21:35.480
speculative. It is not a prediction about real
00:21:35.480 --> 00:21:38.359
extraterrestrial life, and it is not evidence
00:21:38.359 --> 00:21:41.920
for any claim. It is a thought experiment designed
00:21:41.920 --> 00:21:44.359
to make vivid what the process view entails.
00:21:45.160 --> 00:21:48.420
A test of whether the process view can accommodate
00:21:48.420 --> 00:21:51.200
configurations that do not match our familiar
00:21:51.200 --> 00:21:55.589
assumptions. If intelligence is an activity rather
00:21:55.589 --> 00:21:58.670
than a possession, then we should be open to
00:21:58.670 --> 00:22:01.210
finding it in systems that do not resemble our
00:22:01.210 --> 00:22:04.769
own. The question is not whether this alien system
00:22:04.769 --> 00:22:08.119
has a brain. The question is whether it exhibits
00:22:08.119 --> 00:22:11.119
the organized, feedback -sustained, adaptive
00:22:11.119 --> 00:22:14.059
process that the process view identifies with
00:22:14.059 --> 00:22:19.019
intelligence. Where does this leave us? We began
00:22:19.019 --> 00:22:21.799
with a picture of intelligence as a thing inside
00:22:21.799 --> 00:22:25.460
a skull. We have moved through examples that
00:22:25.460 --> 00:22:29.009
complicate that picture. The spider whose sensing
00:22:29.009 --> 00:22:32.660
extends into her web. The person whose calculation
00:22:32.660 --> 00:22:36.400
extends onto paper. The robot whose competence
00:22:36.400 --> 00:22:40.200
extends into its environment. The colony whose
00:22:40.200 --> 00:22:43.180
behavior extends across many interacting members.
00:22:44.099 --> 00:22:46.900
We have considered the proposals of embodied
00:22:46.900 --> 00:22:50.180
cognition and extended mind, treated them as
00:22:50.180 --> 00:22:53.319
live debates rather than settled doctrines. and
00:22:53.319 --> 00:22:55.740
examine the boundary problem that critics raise
00:22:55.740 --> 00:22:59.319
we have seen that the practical question is not
00:22:59.319 --> 00:23:02.579
where the mind really ends but which boundary
00:23:02.579 --> 00:23:05.539
best explains and predicts the behavior we observe
00:23:05.539 --> 00:23:10.759
what emerges is a different picture intelligence
00:23:10.759 --> 00:23:14.559
as a recurring activity sustained by feedback
00:23:14.559 --> 00:23:19.099
distributed across memory body tools other agents
00:23:19.099 --> 00:23:23.170
and an environment It is not that the brain does
00:23:23.170 --> 00:23:26.900
not matter. It is that the brain, for many of
00:23:26.900 --> 00:23:29.960
the behaviors we care about, may not matter alone.
00:23:30.339 --> 00:23:33.940
The process is sustained across time, vulnerable
00:23:33.940 --> 00:23:37.140
to disruption, dependent on stable structures,
00:23:37.400 --> 00:23:39.920
and capable of extending beyond the physical
00:23:39.920 --> 00:23:44.000
boundaries we might naively assume. This is not
00:23:44.000 --> 00:23:48.089
a settled scientific fact. It is a proposal supported
00:23:48.089 --> 00:23:51.890
by evidence contested by serious objections and
00:23:51.890 --> 00:23:55.259
still developing. But it is a proposal that changes
00:23:55.259 --> 00:23:57.779
what we look for when we try to understand how
00:23:57.779 --> 00:24:02.059
thinking works. And that, in the end, is what
00:24:02.059 --> 00:24:06.180
the question was always about. Not a final answer,
00:24:06.400 --> 00:24:08.220
but a better way of asking.