May 29, 2023
Can Machines Exhibit Artificial Anthropocentrism?
I've been thinking about politically
useful euphemisms like "reforestation", and wondering how
LLM based text generators might interact with this common
distortion of language. And what about anthropocentrism?
Since LLM text generators are trained on huge samples of
human generated text, will their output also exhibit
anthropocentrism? How might that work? Perhaps, like human
anthropocentrism, to humanity's long-term detriment. As
with social media, it seems likely that LLM text
generators will get used by some actors to spread and
amplify all sorts of socio-political biases,
with varying levels of subtlety.
Flooding social discussion space with automatically
produced biases would easily drown out attempts at
rational analysis. The increase in "disinformation" (lies,
falsehoods) spread by social media is clear evidence that
enough holes in the boat make bailing unworkable; it has
become too easy to overwhelm the time and energy needed
for rational response. The extreme right has called this
"flooding the zone" - rapidly creating so many lies that
there cannot be enough time to carefully refute them all.
This is in addition to all the other distractions like
"who didn't wear something somewhere". Who with any
decision-making power is asking whether social media has
been positive or negative for society? And what should be
done about it?
Maybe that seems like an insurmountable problem. We
haven't even asked the forest industry to stop calling
clearcuts "harvests". So-called "harvests" that are
supposed to happen every 80-100 years - multiple human
generations - in the future. Our choice of language
matters far more than most people realize. Words like
"efficiency", "productivity", "innovation", "creativity"
and "deregulation" are routinely used in misleading ways
to help further some agenda. This is sometimes called
"spin", which is another euphemism when it really means
witting or unwitting deception. It can be unwitting
self-deception when people uncritically base their
thoughts on misleading language, which is far too common.
Since "forest managers" have plans to convert a
significant proportion of forests to plantations, when
will they be called "plantation managers"? LLM text
generators are unlikely to come up with such a suggestion
- unless prodded by targeted prompts. As critics have
pointed out, the programmed capacity to output acceptable
combinations of words from a massive database is not
intelligence, no matter how convincing that output may
seem. It's unfortunate that machines have been programmed
to mimic language in this way, because humans are already
having a difficult time keeping track of what their use of
language may be doing to themselves. And this may even
mislead some people to think that humans, like these
machines, have also just been programmed to output
combinations of words selected from a massive database. Or
maybe that misunderstanding is one cause of what we're now
facing.
Real human intelligence might help save endangered species
and old growth forests, or prevent wars and pandemics, or
lessen the impact of extreme weather and global heating
from climate change. LLM text generators won't do that.
While they might make a relative few people a lot of money
in the near future, they could easily make many global
problems worse.
What future climate awaits all
these seedlings?
May 22, 2023
Anthropocentrism Again
The photo above shows a lot of tree seedlings in a tree
nursery. These seedlings are a major element in what is
called "reforestation", where seedlings are planted in
recently logged areas. Other elements might include seed
gathering operations, transportation, plantation crews,
and so on. Of course true forests include many other plant
and animal species besides trees. That is just one reason
why you cannot plant a forest. It seems like much more
than a misnomer to call these activities "reforestation".
The result of all the planting will be a plantation.
In its relative uniformity a plantation will lack most of
the qualities of a natural forest. It would take many
human lifetimes for an un-managed plantation to gradually
turn back into an actual forest with complex natural
ecosystems. Of course, forest "managers" are not planning
to let that happen.
The misleading term "reforestation" may seem more
palatable than something crass like "timber cropping" or
"fiber plantation", and could help put curious minds to
sleep. "Reforestation" makes it sound like a forest is
going to be replaced, right? Why bother to wonder whether
an evenly-spaced, single-species block will be equivalent
to a natural forest? It may even be suggested that it will
be a better forest, with genetically improved seed
stock promising faster growing, taller, disease resistant
trees. More wood to cut. Maybe, assuming the climate stays
the same...
It should go without saying that we cannot "improve" a
forest. All our definitions for "good" are based
necessarily on human experience and desires. Even at that,
we very often misjudge what is "good" for ourselves in a
human context. A "better" forest could only mean better
for humans - more useful. It might be important to
recognise this. So we're back to anthropocentrism.
Is anthropocentrism inevitable and unavoidable for humans?
Maybe we could say that there is an innate component and a
learned component. It is certainly possible to recognize
when we are being particularly anthropocentric in our
evaluation of and interaction with other species. If
humans "use" other species wantonly, that could be
considered more anthropocentric than "using" them
conservatively, or even attempting to "preserve" them.
Significantly different interactions - and thought
processes - are possible regardless of whether or not
anthropocentrism is "unavoidable". There is a vast
difference between any self-oriented survival behaviours
in humans and in other species. Yes, all species need to
eat, but that's a little different than wanting to exploit
natural resources to "put bread (and another flat screen
TV) on the table".
How does anthropocentrism influence our thoughts and
perceptions? If we don't know, we will have no idea
whether it is any "good" for humans. Maybe we could start
thinking about that by looking more closely at some of the
euphemisms like "reforestation" that infest our
discussions of "natural resources".
Just another forest fragment
May 15, 2023
Not Seeing the Forest for the Ecosystem Services
I've returned to that forest fragment
along the surging creek, wondering if I can use my
language skills to not see the forest for the "ecosystem
services" it supposedly provides. This time I'm moving
around and not sitting on an uncomfortable rock. Maybe
that will give me a different perspective.
I do appreciate the benefits of so-called "ecosystem
services" - I'm just uncomfortable using those words
thoughtlessly. I appreciate the clear water cascading down
from the melting snow high up in the watershed. The cool
air under the trees is certainly pleasant, in contrast to
the hot sunshine on the road leading to this forest
fragment. Berries and mushrooms for picking will appear in
due course. Ah, but "for picking"? Or water "for
drinking"? Is it really nature's purpose to lay on a
services buffet for us? Is that why we should value nature
- for what it could do for us?
One problem with carving up a whole natural area with
words and attaching quasi-monetary values to the various
bits is that our minds get used to seeing things in
certain ways. We're used to trading bits and pieces too.
The reason it's a forest fragment that I'm in, is
because bits and pieces of the watershed have already been
traded for profits by many independent individuals -
without much thought for cumulative effects on the
interconnections between all the "parts". The water and
cool air are essential for the existence of many life
forms beyond the humans that make use of those qualities
of the forest.
In the real world, people are probably not going to stop
thinking in terms of "natural resources". If you
conceptualise "success" for a species just in terms of
reproduction, humans are probably right up there with
ants. Given the much-vaunted sophistication of the human
mind, it is curious how we are prone to limiting our
thoughts with misnomers. Looking around here in the trees,
I now notice quite a few ants moving to and fro.
Information n'est pas une forĂȘt
May 8, 2023
Information Might Not Be Knowledge
I'm sitting in a forest fragment along
the banks of a creek surging with spring runoff. Somewhere
nearby the long song of a Pacific Wren is clearly audible
even over the din of the cascading water. If I didn't know
how small the wren is, just judging by the volume of the
song I'd be looking all around for a much larger bird. The
song is intricate and pleasant, and free, but is not
likely to be classified as an "ecosystem service".
The water surging down the creek has often been called an
ecosystem service. And so have the trees that have been
cut from the broader watershed and the tourist attractions
of the lakes within that watershed. Here, perched on an
angled boulder, I note that there are no really
comfortable places to sit in sight. So, would that be a
lack of ecosystem services there?
We are so immersed in our anthropocentric viewpoints that
it is little wonder we have trouble thinking deeply about
forests and trees - and all the rest of nature. Some
people even insist that we cannot escape anthropocentrism.
Setting aside the language of "ecosystem services", and
even non-utilitarian anthropocentrisms, what can I see
from my vantage point? Lots of trees and water, obviously.
Lots of trees...or is it really "a lot"? How does the
value judgement implicit in that little phrase, "a lot",
channel our thought processes? Of course trees don't
count, timber does. How about "a lot" of humans? What kind
of an ecosystem service are those? And "a lot" of water
might seem less than enough to a lot of humans wanting to
use it.
The water is moving rapidly and the forest is changing
slowly, so different time scales overlap here. I can't
"see" the long progression of changes that brought the
forest here after the retreat of the glaciers 10.000 years
ago, but I can imagine some bits of it. The ground and
vegetation nearby is quite moist at this time of year.
Tough perennial bracket fungi that are attached to some of
the dying tree trunks and scattered woody debris are
slowly digesting decaying wood. I don't see any mushrooms,
but I know that beneath the visible surface of the ground
mycorrhizal fungi are interconnected with all the tree
roots. I can't see the soil bacteria in their vast,
uncountable numbers, but I know they are there as well.
For that matter, I can't see the trees transpiring. I have
gathered "information" that indicates all these things,
but additional effort is necessary before I can say that I
"know" about them.
What I can "see" depends on what I have
learned and verified - what I "know". Without that
knowledge I wouldn't know where and how to look. Contrary
to "machine learning" tech industry assumptions,
information is only one element in the process of
acquiring knowledge. Information has to be evaluated and
compared to other, prior knowledge and experience. Meaning
isn't inherent in information alone, any more than it is
inherent in individual words. Human interpretation
provides meaning, depending on a complex mixture of
knowledge, belief, and psychological and political
factors. LLM text generators don't understand the
words they have been programmed to generate, and the
humans reading it would be the sole source of any meaning
that the text might be said to have. Meaning can't mean
anything to machines designed to select and output highly
probable combinations of words in response to a prompt -
regardless of how convincing those selections appear to
some uninformed people. Meaning is dynamically negotiated
during ongoing interactions among humans.
Are we starting to miss the forest for the trees here?
Hopefully not. However much you may or may not know about
forests, try sitting quietly in one for a while and notice
as much as you can of your surroundings and your
sensations. Imagine a machine perched awkwardly on an
angled boulder near a surging creek. The machine couldn't
even feel uncomfortable.
Ceci n'est pas une forĂȘt
May 1, 2023
A Bit of Bushwacking
Although the above image of leaves in a
forest might mean many things, it should be obvious that
viewing it is fundamentally different from actually being
in a forest. What can be "known" about a forest from an
image?
It should be obvious that, unlike machines, humans and
other animals come with complex, living, organic bodies.
The presence of microscopic flora and fauna means we can
even say that humans embody ecosystems - and these
ecosystems become part of human minds. The most
complicated mobile robots humans have devised are not
truly comparable. "Artificial Intelligence" and
"Artificial Life" sure seem like curious oxymorons, and
yet those terms are commonly uttered even by people who
should or could know better.
A walk in a forest will vary with the season and weather
conditions and the chosen path - it might even involve
bushwacking. Sights and smells may mean something to each
of us, may impress us in different ways - certainly not in
exactly the same way. People with differing knowledge will
notice different things. Some people might walk in a
forest and not be impressed at all. They might be annoyed,
even. They might think it seems boring or "untidy". That's
life.
Even if a person doesn't know much about forest ecology,
they can certainly enjoy the effects of the forest on
their senses. They might find it calming, or mysterious,
or refreshing in some way. Maybe the textures are
aesthetically pleasing. The enjoyment is a response in
addition to the sensory effects. That's knowledge they
could not get from an image or text. That's also life.
To state the overlooked obvious, living humans change
continuously over time, and so do trees and forests.
Unless we notice new spring growth, or falling leaves, or
a recently fallen tree, we might not notice the slow - by
our standards - changes that forests experience.
Continuous change is only one aspect of forests that an
image or written description cannot really represent.
If you are fortunate enough to walk in a forest and hear
the song of a Pacific Wren, it might bring a smile to your
face. This wren is an unspectacular tiny brown bird with
an amazing ability to project a surprisingly loud song
that is out of all proportion to its size. The birds are
often hard to spot, because their song seems to come from
all directions. Not unlike your perceptions of the forest
itself, if you stop long enough to contemplate it.