Words and phrases

Common words and phrases across the public garden.

Words

Term	Occurrences	Articles
data	1672	49
ai	1462	50
https	787	42
some	538	50
about	377	48
com	374	42
model	373	47
might	349	45
training	322	44
these	316	49
org	315	41
like	290	45
there	286	48
also	278	48
how	269	44
people	268	44
could	264	47
models	253	46
public	242	40
think	241	48
all	234	49
work	231	45
post	228	46
use	223	46
www	221	37
they	215	42
very	213	43
do	211	45
so	206	44
systems	203	40
information	200	33
new	200	43
would	200	41
content	198	31
may	198	43
here	194	42
other	194	45
just	190	42
wikipedia	190	26
one	179	45
see	178	39
make	176	43
many	176	43
should	172	43
want	171	42
out	167	41
knowledge	164	30
collective	160	38
research	156	34
users	154	28
up	153	39
commons	152	23
value	152	35
evaluation	150	23
leverage	148	28
outputs	146	29
good	144	40
open	142	26
even	140	42
into	137	43
openai	135	29
human	132	34
because	131	38
chatgpt	127	27
arxiv	126	33
system	124	32
any	122	40
when	121	38
get	119	37
labor	119	26
set	119	36
my	117	33
etc	113	32
time	112	34
likely	111	38
used	110	37
now	109	32
organizations	109	31
using	109	35
jpg	106	36
re	104	33
much	102	36
still	100	32
most	99	37
code	98	22
help	98	29
kind	98	31
action	95	24
media	95	39
abs	94	31
specific	94	37
while	94	35
making	93	40
need	93	32
blog	92	30
companies	92	25
different	92	33
going	92	34
actually	91	30
high	91	29
social	91	27
coding	90	11
impact	90	31
large	90	38
around	89	35
both	89	35
capabilities	89	26
than	89	35
terms	88	26
case	86	39
something	86	30
world	86	28
agent	85	11
way	85	36
access	84	27
quality	84	23
really	84	32
useful	84	29
important	83	32
platforms	83	20
search	83	24
create	82	31
user	82	19
agents	81	13
certain	81	29
generative	81	21
reddit	80	14
bargaining	79	25
over	79	34
substack	79	20
dataset	78	27
llm	78	24
en	77	25
google	77	26
individual	77	28
between	76	32
output	75	19
strike	74	8
where	74	30
better	73	30
current	73	28
power	73	28
well	73	36
course	72	29
first	72	36
lot	72	27
related	72	25
us	72	29
already	71	27
credit	71	15
ll	71	25
perhaps	71	28
though	71	30
web	71	24
github	70	22
level	70	30
paper	70	24
point	70	37
policy	70	25
rules	70	14
each	69	26
markets	69	12
probably	69	27
no	68	29
support	68	29
approach	67	28
being	67	30
questions	67	27
sharing	67	24
workers	67	12
another	66	31
creators	66	18
part	66	27
right	66	26
source	66	22
tools	66	24
feedback	65	21
points	65	21
same	65	30
wiki	65	25
wikimedia	65	17
labs	64	16
such	64	30
contributions	63	23
does	63	29
economic	63	29
future	63	28
however	63	32
note	63	23
own	63	28
products	63	18
via	63	25
idea	62	30
only	62	29
instance	61	29
market	61	13
say	61	31
term	61	28
back	60	26
dataleverage	60	20
flow	60	21
know	60	30
become	59	30
credits	59	9
down	59	24
general	59	27
incentives	59	18
llms	59	20
particular	59	30
provenance	59	8
through	59	28
two	59	27
across	58	27
short	58	32
try	58	28
above	57	25
based	57	28
file	57	16
key	57	25
then	57	26
towards	57	27
ve	57	24
building	56	20
groups	56	18
me	56	22
consider	55	27
directly	55	28
give	55	26
language	55	21
news	55	23
production	55	21
tech	55	20
them	55	24
end	54	24
issues	54	22
long	54	28
potential	54	24
share	54	21
without	54	29
claude	53	11
contribution	53	21
governance	53	15
learning	53	20
number	53	28
take	53	28
trying	53	29
article	52	26
gpt	52	13
let	52	22
major	52	24
produce	52	27
why	52	25
cases	51	27
provide	51	20
records	51	17
train	51	28
group	50	22
technical	50	22
various	50	24
attestation	49	6
build	49	22
given	49	27
ideas	49	21
legal	49	17
possible	49	24
progress	49	20
times	49	17
ways	49	24
argument	48	24
efforts	48	22
look	48	21
posts	48	22
private	48	14
problem	48	17
rather	48	20
technologies	48	21
unsplash	48	11
benefit	47	22
especially	47	28
full	47	25
process	47	17
against	46	22
datasets	46	20
example	46	20
privacy	46	14
those	46	23
transparency	46	16
writing	46	23
big	45	19
context	45	23
don	45	26
improve	45	19
often	45	25
product	45	15
quite	45	29
similar	45	20
small	45	23
attribution	44	13
discussion	44	24
discussions	44	23

Two-word phrases

Term	Occurrences	Articles
in the	381	50
of the	294	46
https www	210	37
e g	176	37
of data	162	38
it s	154	37
on the	148	41
i think	147	41
to be	138	40
training data	131	37
to the	128	39
want to	127	39
and the	119	35
in a	117	40
the data	109	35
this is	109	37
we can	109	32
the ai	103	34
of ai	99	38
ai systems	97	33
kind of	97	31
there s	96	29
will be	95	34
arxiv org	93	31
https arxiv	93	31
is a	89	37
of a	89	34
org abs	89	30
as a	87	32
can be	87	34
for the	83	34
we should	78	31
from the	76	31
generative ai	76	21
if you	76	22
we might	75	31
to a	74	33
of this	73	32
to make	73	32
there are	72	32
an ai	71	26
set of	71	25
of course	70	29
data leverage	68	17
might be	68	30
public ai	68	9
this post	68	34
a lot	67	26
for a	66	29
terms of	66	25
collective action	63	18
data and	63	29
if we	63	29
would be	63	25
about the	62	30
in this	62	32
going to	61	27
need to	61	29
org wiki	61	24
https dataleverage	60	20
at the	59	30
could be	59	27
is that	59	25
substack com	59	20
the same	59	29
com p	58	19
dataleverage substack	58	19
that the	58	29
a data	56	24
there is	56	30
in some	55	28
to get	54	29
for instance	53	25
some of	52	29
access to	51	23
be a	51	27
in terms	51	20
on a	51	31
from a	50	25
trying to	50	27
with a	50	24
with the	50	27
by the	49	18
have a	49	26
i e	49	27
in which	49	28
lot of	49	24
see e	48	22
wikipedia org	48	18
ai companies	47	14
i m	47	20
of these	47	24
that are	47	24
the case	47	25
try to	47	22
of information	46	15
s a	46	20
the current	46	21
don t	45	26
for ai	45	23
https en	45	18
number of	45	27
en wikipedia	44	18
for data	44	22
to use	44	26
is the	43	27
is to	43	28
that s	43	25
the world	43	20
and data	42	25
we re	41	20
data rules	40	7
to data	40	22
to do	40	26
a model	39	22
all the	39	24
blog post	39	17
coding agents	39	5
data strike	39	6
is not	39	24
open source	39	15
the model	39	20
the public	39	18
ai products	38	13
and ai	38	18
collective bargaining	38	16
may be	37	22
part of	37	23
to see	37	23
you re	37	15
a new	36	22
able to	36	21
about how	36	21
ai models	36	21
data is	36	22
for training	36	17
some kind	36	17
and then	35	22
i ll	35	17
should be	35	20
think the	35	20
a very	34	27
and or	34	21
at least	34	19
the new	34	15
the training	34	18
to ai	34	25
to create	34	18
it is	33	22
that a	33	16
that ai	33	20
that is	33	18
a bit	32	24
data creators	32	15
et al	32	10
in other	32	22
into the	32	16
of people	32	18
openai com	32	18
used to	32	23
and a	31	21
i ve	31	17
that we	31	21
use of	31	19
ai and	30	23
ai is	30	18
and more	30	20
claude code	30	6
coding agent	30	5
language models	30	13
right now	30	18
that can	30	18
to train	30	18
we have	30	18
ai labs	29	11
ai system	29	13
and so	29	18
but i	29	19
data labor	29	13
data protection	29	6
data to	29	15
on data	29	15
people who	29	17
rather than	29	13
that this	29	22
their own	29	19
value of	29	19
we ll	29	15
how the	28	16
let s	28	12
paradox of	28	6
that i	28	17
this will	28	16
what s	28	14
a few	27	19
a way	27	19
acm org	27	18
and i	27	21
case that	27	19
continue to	27	16
data as	27	10
i do	27	14
if the	27	20
much of	27	19
new york	27	8
on this	27	20
org doi	27	19
to build	27	16
training and	27	14
wikimedia commons	27	14
york times	27	8
and evaluation	26	6
be used	26	17
content ecosystems	26	8
data flow	26	13
dl acm	26	18
here is	26	19
https dl	26	18
in many	26	13
like a	26	14
likely to	26	16
make a	26	17
make it	26	15
of reuse	26	6
opt out	26	9
people to	26	15
something like	26	14
the value	26	19
their data	26	15
way to	26	18
we could	26	15
you might	26	15
a large	25	17
ai operators	25	11
are a	25	14
back to	25	12
but it	25	17
by a	25	13
data in	25	17
data pipeworks	25	2
data value	25	8
even if	25	16
frontier ai	25	6
high quality	25	11
https openai	25	16
it seems	25	16
post i	25	21
the future	25	13
a big	24	12
a public	24	10
across the	24	14
because of	24	14
data that	24	16
evaluation data	24	11
for information	24	8
i am	24	15
i d	24	18
i want	24	20
knowledge work	24	6
many of	24	16
markets for	24	7
open knowledge	24	1
the potential	24	14
types of	24	12
work on	24	17
a certain	23	13
ai progress	23	11
and other	23	15
github com	23	14
has been	23	16
in ai	23	16
models and	23	17
other words	23	16
post https	23	16
the long	23	15
the paradox	23	6
think this	23	16
this kind	23	17
to keep	23	16
way that	23	18
we want	23	15
you can	23	13
a good	22	20
about data	22	13
as well	22	17
create a	22	14
data strikes	22	8
end up	22	11
it might	22	14
like the	22	11
model outputs	22	7
note that	22	14
or a	22	13
rely on	22	12
such as	22	15
the more	22	16
the web	22	13
this could	22	15
to say	22	17
users to	22	12

Three-word phrases

Term	Occurrences	Articles
https arxiv org	93	31
arxiv org abs	89	30
dataleverage substack com	58	19
https dataleverage substack	58	19
substack com p	56	19
in terms of	51	20
see e g	48	22
a lot of	45	24
wikipedia org wiki	45	18
en wikipedia org	44	18
https en wikipedia	44	18
there s a	37	17
some kind of	36	17
some of the	33	22
i think the	27	18
new york times	27	8
dl acm org	26	18
https dl acm	26	18
paradox of reuse	26	6
the case that	26	18
acm org doi	25	18
https openai com	25	16
in other words	23	16
this kind of	23	17
i want to	22	18
in a way	22	16
in the long	22	15
the paradox of	22	6
a way that	21	16
https github com	21	12
in this post	21	15
this is a	21	16
and so on	20	15
be able to	20	15
the value of	20	16
https unsplash com	19	11
in some cases	19	14
in the short	19	10
of the data	19	12
the new york	19	8
the training data	19	10
an ai system	18	11
collective bargaining for	18	9
in the ai	18	13
the context of	18	12
the short term	18	10
training data and	18	10
we want to	18	14
a number of	17	14
if you re	17	9
in the context	17	12
in the world	17	11
there is a	17	12
i think it	16	10
the data pipeworks	16	1
the use of	16	13
a variety of	15	14
bargaining for information	15	6
https x com	15	8
i think this	15	10
i think we	15	12
might want to	15	12
the long run	15	10
this post i	15	12
to make a	15	13
algorithmic collective action	14	4
e g the	14	8
for the ai	14	9
kind of data	14	10
live by the	14	5
part of a	14	9
that can be	14	13
the ai field	14	9
think it s	14	11
to be a	14	10
wikimedia org wiki	14	11
evaluation data leverage	13	4
for content ecosystems	13	5
frontier ai auditing	13	2
github com nickmvincent	13	8
i do think	13	8
i think there	13	11
in the same	13	8
it s not	13	11
many of the	13	10
terms of service	13	8
the coding agent	13	3
the fact that	13	10
a data strike	12	4
a kind of	12	8
as part of	12	9
be seen as	12	6
be the case	12	11
but i think	12	12
by the sword	12	4
generative ai systems	12	4
going to be	12	8
https www theverge	12	12
of collective action	12	5
of the ai	12	9
one of the	12	10
openai com index	12	8
the idea that	12	10
to create a	12	9
to try to	12	8
who want to	12	6
www theverge com	12	12
and you are	11	6
are going to	11	7
back to the	11	8
be used to	11	10
commons wikimedia org	11	9
content for training	11	5
data as labor	11	5
e g a	11	9
end of the	11	9
how do we	11	5
https commons wikimedia	11	9
https twitter com	11	8
in the data	11	10
in the future	11	7
it s worth	11	9
let me know	11	8
much of the	11	10
of this post	11	8
of training data	11	8
org wiki file	11	9
points for content	11	5
re going to	11	10
s a lot	11	9
the ai industry	11	6
the end of	11	9
the future of	11	6
the impact of	11	10
the potential for	11	8
tipping points for	11	5
to benefit from	11	6
to get a	11	9
to say that	11	7
we need to	11	8
we should be	11	10
when it comes	11	9
wikimedia commons https	11	11
you are a	11	6
a set of	10	6
ai systems are	10	6
ai technologies are	10	5
are a cartographer	10	5
are system maps	10	5
at some point	10	9
by the free	10	4
free content for	10	4
i ve been	10	8
is awesome and	10	6
it easier to	10	5
make it easier	10	5
maps and you	10	5
of ai systems	10	10
of course there	10	8
of data leverage	10	6
of generative ai	10	7
post https dataleverage	10	8
system maps and	10	5
technologies are system	10	5
the case for	10	6
the case of	10	8
the free content	10	4
the production of	10	8
the right to	10	2
there s no	10	8
there will be	10	8
think we can	10	8
training and evaluation	10	6
a big deal	9	4
a bunch of	9	5
ai output is	9	3
and i think	9	9
and scary you	9	5
any kind of	9	8
at the end	9	7
awesome and scary	9	5
bundles of information	9	4
certain types of	9	5
chatgpt is awesome	9	5
coding agent data	9	2
et al on	9	1
even if you	9	5
if we want	9	8
in the case	9	7
in this space	9	7
it could be	9	7
it may be	9	7
it might be	9	9
it s also	9	8
it would be	9	8
large language models	9	7
of course this	9	8
on the other	9	9
open source software	9	4
openai com blog	9	5
scary you deserve	9	5
terms of data	9	5
the ai supply	9	5
the common crawl	9	4
the creation of	9	8
the number of	9	8
there s also	9	6
this means that	9	8
this post is	9	9
to contribute to	9	9
to engage in	9	7
to opt out	9	4
to think about	9	8
unsplash https unsplash	9	9
used to train	9	8
we re going	9	8
would be a	9	8
you deserve credit	9	5
a data dividend	8	5
a group of	8	8
a large number	8	7
a lot like	8	5
across the ai	8	5
almost everybody including	8	5
along these lines	8	7
and there are	8	7
are likely to	8	7
commons https commons	8	8
d like to	8	8
data provenance initiative	8	4
die by the	8	5
do we know	8	4
everybody including both	8	5
for training data	8	7
google and tiktok	8	5
https www technologyreview	8	8
i do not	8	5
i don t	8	8
i hope to	8	7
i think that	8	6
if you ve	8	7
in order to	8	7
in the past	8	8
including both data	8	5
it comes to	8	7
large number of	8	7
nickmvincent com static	8	7
of ai utility	8	5
of course the	8	8
of the current	8	6
part of the	8	8
some degree of	8	8
stand to benefit	8	5
that in the	8	6
that it s	8	6
that there are	8	8
the ability to	8	7
the ai age	8	5
the benefits of	8	5
the long term	8	7
the other hand	8	8
the potential to	8	4
the process of	8	6
the public ai	8	4
the quality of	8	5
there are many	8	8
there is an	8	8
think there s	8	6
think this is	8	8
this is not	8	7
this will be	8	7
to train on	8	6
to use ai	8	5
to use the	8	7
try to make	8	7
want to know	8	6
www nickmvincent com	8	7
www technologyreview com	8	8
a bit more	7	7
a public ai	7	1
a world in	7	6
agent data deal	7	2
ai supply chain	7	3
along the lines	7	6
an ai product	7	2
and it s	7	7
at a high	7	5
because of the	7	6
but it s	7	6
can be seen	7	5
can help us	7	4
clear data rules	7	4
coding agents are	7	3
collective action with	7	2
com en us	7	4
com p ai	7	6
conscious data contribution	7	4
data creators and	7	3
data e g	7	6
data leverage and	7	6