code
|
text
A Polemic on Terminology
Noah Syrkis
December 29, 2024
In geometry, as laid out by Euclid
[1]
in around 300 BC, it is postulated that
we can:
1.
draw a straight line from any point to any point,
2.
produce a finite straight line continuously in a straight line,
3.
describe a circle with any center and distance (radius),
4.
understand that all right angles are equal to one another,
5.
and finally, understand that if a straight line falling on two straight lines
makes the interior angles on the same side less than two right angles, the two
straight lines, if produced indefinitely, meet on that side on which the angles
are less than two right angles.
For our purposes, this can be simplified to
1.
…. . …….. …. …. … ….. .. … …..,
2.
……. . …… …….. …. …….. …. .. . …….. ….,
3.
…….. . …… …. … …… … …….. (……),
4.
………. …. … ….. …… … ….. .. … …….,
5.
… ……., ………. …. .. . …….. …. ……. .. … …….. ….. …. … …….. …… .. … …. …. …. ….
… ….. ……, … … …… .. ….., .. …….. …………, …. .. …. …. .. ….. … …… … …. …. …
….. …….
We see clearly that the fifth postulate stands out. It is longer (comparable to the
sum of the length of the initial four). It too seems less fundamental and more
convoluted. And yet our intuitions tell us it is a property of geometric spaces
—might it then not follow from the initial four? From the birth of Euclidean
geometry up until the late 18th century, three things were true:
1.
Mathematicians broadly thought the postulate a necessary feature of geom
etry.
2.
Mathematicians broadly thought the postulate could be derived from the first
four.
3.
Mathematicians were unable to derive it from the first four.
fifth shows the fifth postulate visually:
if two lines are not parallel, they will
eventually cross exactly once; in our case a few centimeters to the right of where
the figure ends.
Only in 1872 was the problem finally resolved by the Erlangen program
[2]
.
The fifth postulate, rather than being a true property of geometry that we were
so far unable to prove, was but a postulate of a particular
kind
of geometry,
now known as Euclidean geometry. Geometry, however, is much broader than
that. The Erlangen program paved the way for non-Euclidean geometry, famous
examples of which are hyperbolic geometry and elliptic geometry, both of which
follow from varying the fifth postulate, and are in broad use today, a prerequisite
for much of physics, engineering, math, etc. Non-Euclidean geometries surround
us: we live on a sphere, many conceptual spaces seem to bend in all kinds of
strange ways. Indeed,
M. M. Bronstein, J. Bruna, T. Cohen, and P. Velǐcković [3]
introduces geometry deep learning. So why the multi-millennia-long confusion?
It has been argued
[4]
that the use of the terms “line” and “point” in Euclid’s
postulates were so pregnant with association from our real-world experience,
that we implicitly, without knowing so, attributed qualities to these mathemat
ical constructs. If using the terms “line” and “point” will lead humanity on a
two-millennia-long treasure hunt, what might terms like “artificial intelligence”,
“neural network”, and “evolutionary computation” do to our clarity of thought?
These terms are obviously sexy to say. The terms are sure to have played
essential roles in fundraising, product pitching, PhD applications, and even the
occasional one-night stand. But alas if the price of unnoted confusion is a halt to
progress, something must be done. Clarity of thought is a core value of the Virian
Project, and thus, it is in this spirit that we produce the dictionary for clarity of
thought seen in
Table 1
. If nukes, as some have said, are viagra for politicians
[5]
, so too the use of these overly smart sounding terms is for the scientist. You,
dear reader to the extent that the profession applies, are encouraged to use these
terms in academic writing, but fear not, for one-night stands and fundraising,
pop a pill, and go from the 2nd to 1st column.
Viagra term
Virian term
neural network
layered filter
deep learning
layered filtering
language model
text model
foundation model
modality model
evolutionary computa
tion
genetic programming
artificial intelligence
machine learning
References
[1]
Euclides, T. L. Heath, D. Densmore, and Euclides,
Euclid's Elements: All
Thirteen Books Complete in One Volume
. Santa Fe, NM: Green Lion Press,
2007.
[2]
F. Klein,
Elementary Mathematics from an Advanced Standpoint. Geometry
,
Dover edition. Mineola, N.Y: Dover Publications, 2004.
[3]
M. M. Bronstein, J. Bruna, T. Cohen, and P. Velǐcković, “Geometric
Deep Learning: Grids, Groups, Graphs, Geodesics, and Gauges,” no.
arXiv:2104.13478. arXiv, May 2021. doi:
10.48550/arXiv.2104.13478
.
[4]
D. R. Hofstadter,
Gödel, Escher, Bach: An Eternal Golden Braid
, 20th-anniver
sary ed. London: Penguin, 2000.
[5]
A. Sirkis,
Descarbonário
. Ubook Editora, 2020.