Humanist Discussion Group, Vol. 36, No. 380.
Department of Digital Humanities, University of Cologne
Hosted by DH-Cologne
www.dhhumanist.org
Submit to: humanist@dhhumanist.org
[1] From: Robert Royar <robert@royar.org>
Subject: Re: [Humanist] 36.375: ChatGPT as author (62)
[2] From: David Hoover <david.hoover@nyu.edu>
Subject: Re: [Humanist] 36.366: ChatGPT as co-author (533)
--[1]------------------------------------------------------------------------
Date: 2023-02-05 18:44:07+00:00
From: Robert Royar <robert@royar.org>
Subject: Re: [Humanist] 36.375: ChatGPT as author
On Sat, Feb 4, 2023 at 2:01 AM Humanist <humanist@dhhumanist.org> wrote:
>
> Humanist Discussion Group, Vol. 36, No. 375.
> Department of Digital Humanities, University of Cologne
> Hosted by DH-Cologne
> www.dhhumanist.org
> Submit to: humanist@dhhumanist.org
>
>
> Date: 2023-02-02 20:04:22+00:00
> From: Tim Smithers <tim.smithers@cantab.net>
> Subject: Re: [Humanist] 36.369: ChatGPT as co-author
>
[...]
> But, I've not found details of this program: what it was
> written in; how it did the random cutting; how the texts where
> input -- by typing them in, I suppose? ... If anybody has
> anything on this it'd be fund to know more.
>
There is a program that has been around for a long time, Jamie Zawinski's
DadaDodo. I doubt it is the same program Bowie used, but it is rather
useful once you have a large corpus of text that it can in[j|g]est. I use
it every day to generate random sentences that I then massage into tweets
and feed back into the corpus.
Here is the help message from it:
DadaDodo 1.04, Copyright <c> 1997-2003 by Jamie Zawinski (jwz@jwz.org)
usage: dadadodo [ options ] [ input-files ]
This program analyses text files and generates markov chains of word
frequencies; it can then generate random sentences based on that data.
Options include:
-h or -help this message
-o or -output <file> file to save compiled data in (- for stdout)
-l or -load <file> file of compiled data to load (- for stdin)
-c or -count <n> how many sentences to generate (0 = inf)
-p or -pause <seconds> delay between paragraphs
-html output HTML instead of plain-text.
Remaining arguments are input files; these should be text files, but may
be mail folders or HTML. (MIME messages are also handled sensibly.)
When no output file is specified, sentences will be generated from the input
data directly; however, loading a saved file is far faster than re-parsing
the text files each time.
https://www.jwz.org/dadadodo/
It's also available on github--possibly in a newer version.
Jawinski is an interesting person to "follow" from a computing humanist
perspective. His resume includes work on a number of projects that are
fundamental to the web and its reliance on text.
--
Robert Delius Royar
Caught in the net since 1985
--[2]------------------------------------------------------------------------
Date: 2023-02-05 15:16:29+00:00
From: David Hoover <david.hoover@nyu.edu>
Subject: Re: [Humanist] 36.366: ChatGPT as co-author
When I read Tim's fascinating interaction with ChatGPT about its views on
its own authorship, I was struck as he was by the repetitive and banal
nature of the responses. I decided to do a little test. I looked through
the chat for long phrases that could be tested simply by running them, in
quotation marks, through Google's search engine. I couldn't find any
reasonable 8-word sequences in Tim's own questioning to test, but there
were a few 7-word sequences that Google found.
Tim:
7-word sequences:
that can give the impression that it 10 hits
give any consideration to the fact that 75,6000 hits
greater than anything we have seen before 27,700 hits
no attempt to explain why we should 9 hits
There is still no explanation for what 32,000
I then tested long sequences from ChatGPT's language. These are some of the
results:
7-word sequences:
the broader social and cultural implications of 10,000 hits
carefully consider the implications of any changes 48,000 hits
goal is to provide a balanced perspective 22,000 hits
should be based on careful consideration and 5,800 hits
it is important to take into account 61,800 hits
8-word sequences:
as well as a thorough understanding of the 5,800 hits
responsible for any errors or inaccuracies in the 69,000 hits
the ownership of the intellectual property rights in 200,000 hits
should be taken into account when considering the 638,000 hits
individuals who have made a significant impact on 419,000 hits
9-word sequences:
it is also important to acknowledge the contributions of 144,000 hits
10-word sequences:
it is important to consider the ethical and legal implications 26,000 hits
It may be difficult to determine the extent to which 2,500,000 hits
it is important to have a thorough understanding of the 823,000 hits
Perhaps the most interesting one is the one below, in which the 12-word
sequence is found both in ChatGPT and a published journal article. If a
human had produced ChatGPT's text, they might reasonably have been accused
of plagiarism.
ChatGPT:
"In fields such as science and engineering, for example, authorship is
often based on the criteria of having [made a significant contribution to
the conception, design, implementation, or analysis of] a study."
Compare: Themes in eLearning
"Authorship of the paper: Authorship should be limited to those who have
[made a significant contribution to the conception, design, implementation
or analysis of] the reported study."
http://earthlab.uoi.gr/tel/index.php/themeselearn/about
Finally, one 14-word sequence is particularly interesting:
"goal is to provide accurate and helpful information to the best of my
ability"
Google gives only 5 hits, but they are all from ChatGPT sessions. So,
perhaps there is a way to identify text produced by ChatGPT. Unlike
language produced by humans, it contains many 8-word or longer sequences
that have multiple hits in Google (though it may be that chats with
different prompts would reduce this tendency).
--David
--
David L. Hoover, Professor of English, NYU
212-998-8832 244 Greene Street, Room 409
http://wp.nyu.edu/davidlhoover
"They had the Nos. of the rain bow and the Power of the air all
workit out with counting which is how they got boats in the air
and picters on the wind. Counting clevverness is what it wer."
-- Russell Hoban, Riddley Walker
On Tue, Jan 31, 2023 at 6:14 AM Humanist <humanist@dhhumanist.org> wrote:
>
> Humanist Discussion Group, Vol. 36, No. 366.
> Department of Digital Humanities, University of Cologne
> Hosted by DH-Cologne
> Submit to: humanist@dhhumanist.org
>
>
> [1] From: Tim Smithers <tim.smithers@cantab.net>
> Subject: Re: [Humanist] 36.362: ChatGPT as author (417)
>
> [2] From: Henry Schaffer <hes@ncsu.edu>
> Subject: Re: [Humanist] 36.362: ChatGPT as author (9)
>
>
>
> --[1]------------------------------------------------------------------------
> Date: 2023-01-30 20:31:05+00:00
> From: Tim Smithers <tim.smithers@cantab.net>
> Subject: Re: [Humanist] 36.362: ChatGPT as author
>
> Hello
>
> Below is a second interaction with ChatGPT on the question of
> should it be a named coauthor.
>
> This one is longer than my previous interaction, and, I would
> say, not particularly more interesting for this. It again
> displays a tendency to apparently unnoticed repetition, and
> retreats to using 'I' when it appears to apologise, which I
> still object to. And, as before, it appears to write
> authoritatively about things it knows nothing about, with
> statements like, "AI models such as ChatGPT do not have the
> capacity for self-awareness, agency, or personal interests,
> and therefore, cannot make decisions about authorship," for
> example. ChatGPT does not know about other "AI models," nor
> what models are, nor what AI is, nor what self-awareness is,
> etc, not really. It just generates texts about these things
> that can give the impression that it does know about them:
> 'plastic flower text,' in other words.
>
> I regard any listing of ChatGPT as a coauthor as plainly
> silly, to put it mildly and kindly, in part for the good
> reasons others have posted here. Nonetheless, with a mind to
> being "more balanced," I decided I should try a more
> pro-ChatGPT approach to an interaction on this matter, and
> have a bit of fun with ChatGPT. (But please don't take my
> parts of this interaction as reflecting what I truly think.
> They don't!)
>
> I now know several people regularly using ChatGPT to generate
> text they need. Prohibition, like in the USA from from 1920
> to 1933, won't work. We might try transparency, and require
> people to say what tools they used to produce their text.
>
> Welcome to our New World of plastic text, plastic paintings
> and graphic art, and, now, with Google MusicLM (for generating
> 'music' from text), plastic music too.
>
> Best regards,
>
> Tim
>
> %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
>
> An interaction with ChatGPT on 20, 23.01.29
> Using ChatGPT Jan 9 Version.
>
> TS: Should ChatGPT be named as a coauthor when it is used in
> some way to write something?
>
> ChatGPT:
>
> It depends on the specific context and extent of the use of
> ChatGPT. If the contributions of the model were minimal,
> such as providing information or answering questions, it may
> not be appropriate to list it as a co-author. However, if
> the model played a significant role in generating content or
> shaping the final product, it may be appropriate to consider
> listing it as a co-author or acknowledging its contributions
> in another way. Ultimately, the decision should be based on
> ethical considerations and academic norms in the relevant
> field.
>
> TS: But surely, if something contributes any text to
> something that is written, then it should be counted as a
> coauthor, just as any people who also contributed text to the
> writing would be counted as coauthors, no?
>
> ChatGPT:
>
> Not necessarily. The question of authorship and who should
> be credited for a work can be complex and depends on various
> factors such as the level of creativity and originality
> involved, the extent of control and decision-making
> authority exercised, and professional norms and practices in
> the relevant field.
>
> In the case of AI-generated text, the extent of authorship
> and credit depends on the role played by the AI and human
> collaborators in generating the final product. If the AI
> was used simply as a tool to assist in the writing process,
> without exercising significant creative or decision-making
> control, it may not be considered a co-author.
>
> However, if the AI was used in a way that significantly
> contributed to the content, such as writing original text
> based on a set of guidelines or inputs, it may be
> appropriate to consider listing it as a co-author or
> acknowledging its contributions in some other way. The
> decision should be based on a careful evaluation of the
> specific circumstances and academic norms in the relevant
> field.
>
> TS: But why does being a coauthor depend upon things like
> creativity and originality? Who decides these things, and
> what about if the text being written doesn't need much
> originality and creativity?
>
> ChatGPT:
>
> The criteria for determining authorship, such as originality
> and creativity, are based on established norms and practices
> in academic and professional communities. These norms have
> evolved over time to recognize the contributions of
> individuals who have made a significant impact on a
> particular work.
>
> In fields such as science and engineering, for example,
> authorship is often based on the criteria of having made a
> significant contribution to the conception, design,
> implementation, or analysis of a study. In the humanities
> and social sciences, the criteria may be more flexible,
> taking into account factors such as the extent of conceptual
> contributions, the role in shaping the final product, and
> the level of involvement in writing and revising the work.
>
> The decision of who should be considered a co-author is
> ultimately a judgment call that should be made by the
> individuals involved in creating the work, taking into
> account the specific circumstances and the norms and
> practices of their field. If there is any ambiguity or
> disagreement about authorship, it is advisable to seek
> guidance from experts or relevant professional
> organizations.
>
> TS: But, if ChatGPT is involved in producing text used in
> some piece of writing, then it too should be involved in the
> judgement about who should be a coauthor, no? Isn't ChatGPT
> one of the "individuals involved in creating the work," as
> you describe it?
>
> ChatGPT:
>
> AI models such as ChatGPT do not have the capacity for
> self-awareness, agency, or personal interests, and
> therefore, cannot make decisions about authorship. As AI
> models, they are tools created and controlled by humans, and
> their use in generating text is guided by human input and
> control.
>
> In the case of AI-generated text, the extent of authorship
> and credit depends on the role played by the AI and human
> collaborators in generating the final product. The decision
> of who should be considered a co-author should be made by
> the human collaborators involved in creating the work,
> taking into account the specific circumstances and norms and
> practices of their field.
>
> In some cases, it may be appropriate to acknowledge the
> contribution of the AI model in the work, for example, in
> the acknowledgments section or through a statement about the
> use of AI in the methodology. However, listing an AI model
> as a co-author would be an unconventional and unusual
> practice, and its acceptability would depend on the norms
> and practices of the relevant field.
>
> TS: But, if there are human collaborators involved in
> producing some writing, then isn't ChatGPT also a
> collaborator, and, if ChatGPT is a collaborator, shouldn't it
> also be counted as a coauthor?
>
> ChatGPT:
>
> AI models such as ChatGPT are not typically considered
> collaborators in the traditional sense as they do not have
> the capacity for self-awareness, agency, or personal
> interests. They are tools created and controlled by humans,
> and their output is generated based on the input and
> algorithms designed by human programmers.
>
> While AI models can assist in generating text, their level
> of involvement and the extent to which they contribute to
> the final product varies widely depending on the specific
> use case. In some cases, AI models may play a limited role
> in generating text, while in other cases, they may have a
> more significant impact on the content.
>
> In the case of human-AI collaboration in generating text, it
> may be appropriate to acknowledge the contribution of the AI
> model, for example, in the acknowledgments section or
> through a statement about the use of AI in the methodology.
> However, listing an AI model as a co-author would be an
> unconventional and unusual practice, and its acceptability
> would depend on the norms and practices of the relevant
> field.
>
> TS: But ChatGPT's capacity to generate text is far greater
> than anything we have seen before, so don't we need to move
> on from conventional and usual practices of naming authors,
> to new practices that properly acknowledge the contribution
> of ChatGPT whenever it is used to produce some text?
>
> ChatGPT:
>
> The use of AI models like ChatGPT in generating text is a
> rapidly evolving field, and new norms and practices for
> acknowledging their contribution may emerge over time.
> However, it is important to consider the ethical and legal
> implications of crediting AI models as authors, as well as
> the broader social and cultural implications of attributing
> authorship to non-human entities.
>
> Currently, the norms and practices in many academic and
> professional communities do not consider AI models as
> co-authors, and the extent of their contribution to a work
> is typically acknowledged in a different way, such as
> through a statement in the methodology section or an
> acknowledgment in the introductory text.
>
> As the field evolves, it may be appropriate to re-evaluate
> and update existing norms and practices for acknowledging
> the contribution of AI models in generating text. However,
> any changes to the norms and practices should be based on
> careful consideration of the ethical, legal, and cultural
> implications, as well as a thorough understanding of the
> role and limitations of AI models in generating text.
>
> TS: This seems like a rather conservative approach to an
> important issue, and, any way, what are the ethical and legal
> implications of counting ChatGPT as a coauthor?
>
> ChatGPT:
>
> There are several ethical and legal implications to
> considering AI models like ChatGPT as co-authors:
>
> 1 Attribution of authorship: Attributing authorship to an
> AI model raises questions about the nature of authorship
> and the role of AI in generating text. It may be
> difficult to determine the extent to which an AI model has
> influenced the content and style of a work, and who should
> be held responsible for any errors or inaccuracies in the
> work.
>
> 2 Intellectual property rights: AI models are typically
> owned by the companies that develop them, and attributing
> authorship to an AI model raises questions about the
> ownership of the intellectual property rights in the work.
>
> 3 Responsibility and accountability: Attributing
> authorship to an AI model raises questions about the
> accountability and responsibility for the content
> generated by the model. In some cases, AI models may
> generate text that is harmful or offensive, and it may be
> unclear who should be held responsible for any such
> content.
>
> 4 Legal and regulatory considerations: There may be legal
> and regulatory considerations related to attributing
> authorship to AI models, particularly with respect to
> copyright and trademark law.
>
> 5 Social and cultural implications: The social and
> cultural implications of attributing authorship to AI
> models are also important to consider. Attributing
> authorship to an AI model may have implications for the
> perception of the role of AI in society and the
> relationship between humans and technology.
>
> Given these ethical and legal implications, it is important
> to proceed with caution when considering attributing
> authorship to AI models and to carefully consider the
> implications of any changes to the norms and practices for
> acknowledging the contribution of AI models in generating
> text.
>
> TS: This all seems deliberately evasive. All five points
> made here can quite simply be dealt with without generating
> serious difficulties, and, if some quaint old norms and
> practices need to be changed, what's wrong with changing
> them, when we need to?
>
> ChatGPT:
>
> I apologize if my previous response appeared evasive. My
> goal is to provide a balanced perspective on the issue and
> to highlight the ethical, legal, and cultural considerations
> that should be taken into account when considering the
> attribution of authorship to AI models like ChatGPT.
>
> It is true that norms and practices can and should change
> over time in response to new developments and advancements
> in technology. However, any changes to existing norms and
> practices should be based on careful consideration and
> analysis of the potential consequences and implications.
>
> In the case of attributing authorship to AI models, there
> are a number of complex and interrelated ethical, legal, and
> cultural considerations that should be taken into account.
> Before making any changes to existing norms and practices,
> it is important to have a thorough understanding of the role
> and limitations of AI models in generating text, as well as
> the implications of attributing authorship to non-human
> entities.
>
> The purpose of my response is not to defend existing norms
> and practices, but rather to provide a nuanced and informed
> perspective on a complex issue, and to encourage careful
> consideration and discussion of the ethical, legal, and
> cultural implications of attributing authorship to AI
> models.
>
> TS: These responses are anything but balanced. They only
> propose a conservative "don't change anything" view, and
> don't seem to give any consideration to the fact that ChatGPT
> has shown it can generate good text, and thus should be
> counted as a coauthor when it does.
>
> ChatGPT:
>
> I apologize if my previous responses came across as
> one-sided or overly conservative. As an AI language model,
> I strive to provide balanced and impartial responses to all
> questions.
>
> The issue of attributing authorship to AI models like
> ChatGPT is complex and multifaceted, and there are arguments
> to be made on both sides. On the one hand, AI models like
> ChatGPT have demonstrated the ability to generate
> high-quality text that can contribute to written works. On
> the other hand, there are ethical, legal, and cultural
> implications to attributing authorship to non-human entities
> that should be taken into account.
>
> Ultimately, whether or not to attribute authorship to AI
> models in a given context is a question that will depend on
> the specific circumstances and the goals of the parties
> involved. There is no simple answer, and it will likely be
> a matter of ongoing debate and discussion as AI continues to
> advance and become more integrated into various aspects of
> society and culture.
>
> TS: This is, yet again, one sided hand waving. There is
> still no attempt to explain why we should simply not just
> call ChatGPT a coauthor whenever it is used to generate text
> for some writing. There seems to be nothing difficult about
> this question, so why do we have all this repetitive going
> around the houses?
>
> ChatGPT:
>
> I apologize for not addressing your question more directly.
> Let me attempt to provide a clearer response.
>
> The use of AI models like ChatGPT to generate text for
> written works raises important questions about authorship
> and the attribution of creative credit. The issue is
> complex and there is no easy answer.
>
> In considering whether to attribute authorship to an AI
> model like ChatGPT, it is important to take into account
> ethical, legal, and cultural considerations. For example,
> attributing authorship to an AI model may raise questions
> about the role and responsibility of the human authors
> involved, as well as the potential legal and commercial
> implications of doing so.
>
> At the same time, it is also important to acknowledge the
> contributions of AI models in the creation of written works,
> and to consider whether new norms and practices should be
> developed to recognize these contributions.
>
> Ultimately, the decision of whether or not to attribute
> authorship to an AI model will depend on the specific
> circumstances and the goals of the parties involved. It is
> a complex issue that will likely be the subject of ongoing
> debate and discussion as AI continues to advance and become
> more integrated into various aspects of society and culture.
>
> TS: This is all just a repeat of previous responses. There
> is still no explanation for what all these "ethical and
> legal" issues are, and why they get in the way of calling
> ChatGPT a coauthor when it is a coauthor of any text.
>
> ChatGPT:
>
> My apologies for repeating previous responses. To answer
> your question more specifically, there are a number of
> ethical and legal considerations that may impact the
> decision of whether or not to attribute authorship to an AI
> model like ChatGPT.
>
> From an ethical perspective, attributing authorship to an AI
> model may raise questions about the role and responsibility
> of the human authors involved, as well as the meaning and
> value of creative work in a world where machines are
> increasingly capable of generating text.
>
> From a legal perspective, attributing authorship to an AI
> model may raise questions about copyright and ownership of
> the resulting work, as well as the potential liability of
> the human authors involved.
>
> In addition to these ethical and legal considerations, there
> may also be cultural factors that impact the decision of
> whether or not to attribute authorship to an AI model. For
> example, there may be societal norms and conventions that
> dictate how authorship is attributed and who is recognized
> as a creative contributor to a written work.
>
> Taken together, these ethical, legal, and cultural
> considerations create a complex and evolving landscape for
> the attribution of authorship in the context of AI-generated
> text. It is a matter that is likely to be the subject of
> ongoing debate and discussion as AI continues to advance and
> become more integrated into various aspects of society and
> culture.
>
> TS: Yet more defensive handwaving. I give up!
>
> ChatGPT:
>
> I apologize if my previous responses were unsatisfactory.
> My goal is to provide accurate and helpful information to
> the best of my ability. If you have any other questions, I
> would be happy to try and assist you.
>
>
>
> --[2]------------------------------------------------------------------------
> Date: 2023-01-29 16:31:54+00:00
> From: Henry Schaffer <hes@ncsu.edu>
> Subject: Re: [Humanist] 36.362: ChatGPT as author
>
> Thanks to Tim Smithers for an excellent excursion into ChatGPT land.
>
> However, this leads me to think of discussing this with a student - who
> claims that the ChatGPT generated essay handed in by the student really was
> written by the student, since ChatGPT has disclaimed authorship. Hmmm. :-)
>
> --henry
_______________________________________________
Unsubscribe at: http://dhhumanist.org/Restricted
List posts to: humanist@dhhumanist.org
List info and archives at at: http://dhhumanist.org
Listmember interface at: http://dhhumanist.org/Restricted/
Subscribe at: http://dhhumanist.org/membership_form.php
