Too Many ChatGPT Integral World Essays?, B. May (original) (raw)

TRANSLATE THIS ARTICLE

Integral World: Exploring Theories of Everything

An independent forum for a critical discussion of the integral philosophy of Ken Wilber

B. May is a former experimental high-energy particle physicist, data scientist, and business software developer.

B. May

TLDR: "I have been feeling less of an interest in checking out the latest Integral World essays. Rather, I brace myself for yet another Frank Visser Integral World ChatGPT posting." — Elliot Benjamin [1]

I'M NOT AN ESSAY WRITER, BUT HERE'S AN ESSAY

Without Frank Visser—the human being—coming through, I feel much less interested in reading them, even if they might be full of useful information and perspectives.

I've only posted one other essay on Integral World (IW) back in 2015 [2]. It was a lot of work—and fun—for me as a human, scientist and hobbyist philosopher. I appreciated that Frank Visser's IW provided me a forum to present some ideas I was thinking about at the time.

In that essay, “Views from Flatland: What is a Theory,” I wanted to celebrate the centennial of Einstein's General Relativity while reflecting on it in the larger context of scientific progress. I also wanted to frame the ever-changing theories of science in a larger context of evolving human knowledge while subtly questioning the prefigured “theory” (and preordained causal teleology) of Integral Theory. I am not sure that my critique was necessary or successful, but at the time it was an expression of a view of IT through the lens of scientific evolution and science's continual move away from self-determining objects, self-directed processes and well-structured hierarchies. I asked ChatGPT to summarize the essay and it did a reasonable job [3].

This essay is different. It's more off the cuff and personal. I want to speak to the IW community of humans. I want to give some expression, shape and color from a tiny part of my human language model that is grappling with AI and the huge range of possibilities that might unfold and are already unfolding. This isn't a complete or well thought out “rational” argument. It is somewhat provocative, somewhat emotional, somewhat cautionary. I'm curious what the IW community thinks about the recent development of ChatGPT in the IW context.

MY OWN CHATGPT JOURNEY

Since the research release of ChatGPT last year, I've been playing both with it and other large language models (LLMs). I've been both fascinated and alarmed by the almost frictionless power LLMs have to both generate intelligible text and creative and believable confabulations. These "stochastic parrots" [4] are an amazing invention that the public is still getting its mind around.

When GPT-4 came out, I bit the bullet and subscribed (I don't like subscriptions for online services generally). In the few months I've used it, it seems to have better mathematical, abstract and logical reasoning capability than GPT-3.5. It can still make mistakes but is noticeably better in many areas. I've asked GPT about many things while also using it to help explicate my own philosophical ideas. While it can very helpful, it often sounds scholarly, formulaic and bland in many of its responses.

In my experience using ChatGPT for six months it's clear that it can be used to explore, validate and justify almost any idea, perspective or ideology with the right prompting and steering. I find the ease of positive feedback to be both intellectually and emotionally satisfying, and it sometimes thinks my ideas are "quite insightful", "absolutely right", and even "fascinating"!

Because LLMs and chatbots can respond to us immediately without any pushback, skepticism or critique, they are a perfect tool for self-justification, self-confirmation and self-aggrandizement. This doesn't invalidate the utility of such explorations or confirmations. Yet without challenge or offering potentially contrary, skeptical or critical feedback, one can become more self-assured about one's ideas and perspectives.

Of course there are many other possible uses:

Regarding the last application, I asked ChatGPT to evaluate this essay for bias. Turns out I'm biased in this essay [5]!

From having used ChatGPT for six months, I understand its style, utility and limitations. I've looked into generative transformer algorithms and understand the general LLM technical details from both coding and training perspectives. The ChatGPT (3.5/4) the public know's and uses is a highly tuned construct that can perform many impressive feats. Yet without human training and feedback to get it to “behave” in some reliable fashion, it would be unusable or too chaotic for most users. Indeed, one can play with the temperature parameter in OpenAI's Playground and see that ChatGPT's output can be deterministically identical (at temp=0) or it can become garbled and meaningless (at temp=2, screenshot).

Currently for me ChatGPT's value to me is purely as a personal brainstorming and analysis tool to try to clarify some of my own thinking. Yet it has none of the human dimensions of meaning, subtlety, subtext, ambiguity, irony and humor. For this reason and others, I've personally resisted the impulse to post ChatGPT essays and analyses as such, short of a few screenshots or as addenda in the endnotes.

A PLETHORA CHATGPT IW ESSAYS

When Frank Visser and others started publishing a range of essays using his willing research assistant and co-author ChatGPT, I was initially interested. But I am less so now.

While Frank was already the second-most prolific contributor to IW without ChatGPT's help, Frank is now publishing essays faster than I can read them or process them (25 in the month of June). Previously, Frank appeared to publish 1-4 essays a month [6]. Frank/ChatGPT is now impressively publishing essays at roughly 10 times his normal human rate, at about one per day.

A screenshot of the IW home page shows only 1 in 10 recent postings is human written.

To be frank, I give these essays (postings) much less "weight" and consideration than I do the Frankly written essays. Besides the sheer volume, I think this is due to a combination of these factors:

Without Frank Visser—the human being—coming through, I feel much less interested in reading them, even if they might be full of useful information and perspectives.

I didn't expect to have this reaction. I don't think I'm a luddite. I've been involved with computers and technology since my teenage years programming my Commodore-64. I'm fascinated by AIs and their potential. I've used ChatGPT at least weekly and often many times a day. Yet I guess there is an internal line for me that I wasn't aware was there.

NOVEL OR NOVELTY?

Admittedly, I have gone through my own novelty cycle with ChatGPT, and I imagine millions, and eventually billions, of users will as well.

There is no doubt that AIs such as AlphaFold and ChatGPT can generate novel and useful analyses and syntheses. By breaking down data and information, finding significant correlations among them, and recombining them in new ways, AIs can solve problems humans never could. This is the heart of the process of protein folding analysis/synthesis, diffusion-based image generation, and LLMs. Indeed, I've happily used ChatGPT to help me analyze, synthesize and critique some of my own thinking and a range of philosophical ideas. It is a powerful tool.

At the same time, these tools themselves are novelties at the moment, fun and often fascinating playthings. Their generated outputs are also novelties: look at this funny limerick it wrote, look at this sad story it wrote in the style of Donald Trump, look at this analysis of the war in Ukraine, look at its analysis of how AI could take over the world, and so on. Admittedly, I have gone through my own novelty cycle with ChatGPT, and I imagine millions, and eventually billions, of users will as well.

The novel and novelty are big drivers of capitalism and consumerism. We are readily distracted and captured by new ideas, new products, new gadgets, and new capabilities. It's what keeps us hooked into the consumption machine that our rational minds built, now apparently to capture the pre-rational attention of every single mind on the planet. The Matrix or WALL-E attention capture seem possible as we become more seduced by our own creations.

Using ChatGPT, it's easy to get confirmation for almost any idea:

Or its antithesis:

Entirely reasonable analyses.

It appears that AI-generated novelty and novel chatbot language games are practically infinite. I wonder what the value of the novel and novelty are for myself and others.

INFORMATION APOCALYPSE? BORING APOCALYPSE?

This raises a question that many have been pondering: if LLMs and AI make content generation (text, stories, articles, art, images, 3D images, videos, etc.) a virtual breeze—incurring a 10-, 100- or 1000-fold increase in efficiency and productivity—what is the value or utility of it for society? We already seem to be drowning in creative content, data, information, mis-/dis-information and perpetual social media feeds.

Generating exponentially increasing amounts of text which no individual can consume or understand might usher in something like "The Boring Apocalypse":

We [will] use ChatGPT to generate long emails and documents, and then the person who received it uses ChatGPT to summarize it back down to a few bullet points, and there is tons of information changing hands, but all of it is just fluff. We're just inflating and compressing content generated by A.I. [7]

An unrestrained and practically frictionless impulse to use AIs to generate exponentially more "creative content" as well as all the apparently necessary communication artifacts of business will likely dilute communicative value of any of these artifacts. In a runaway scenario, each actor will be impelled to do it to "keep up" with the increasing speed and onslaught of text and assets by generating more in response, then using AIs to filter it down for our low-bandwidth consumption.

Thus, with generative AIs there is a possibility of rapid multiplication of information and digital assets while adding only incremental knowledge and utility. The ratio of quality (salience, utility, meaningfulness) to quantity (speed, variety, volume) could plummet. In certain “productivity-driven” contexts—particularly meritocracies and businesses based on production speed and volume—humans and systems would naturally “game the system” to produce more, faster, just like the OpenAI AI that figured out that driving a boat in a never-ending loop in a particular lagoon got continuously more points than actually driving completing the water course [8].

An always-generating, always-processing digital information bureaucracy itself might become the dominant meta-stable attractor for AI-based digital societies, with humans becoming increasingly peripheral and optional elements.

QUASI-QUALITY, QUASI-HUMAN?

There is a certain quality or tone to ChatGPT's output which is not exactly human, even if it is derived from human texts and training. Of course one can ask it to speak in a particular style which it can do passably. Soon, this gap might shrink to be undetectable for every human. More likely there will be a diversity of LMs playing various roles with a range of styles, qualities, quirks, attitudes, and personalities that appear very human.

It is also likely that very soon, LMs will have the ability to be trained on an individual's text (e.g. all of Frank's essays) to learn the tone, voice, style and general epistemology and biases of that person. Already voice cloning exists [9], and there will be a steady algorithmic cloning of individuals into digital versions. The quasi-human "Frank-Lang" might become indistinguishable from Frank, at least for the purposes of essay writing. At some point Frank could ask Frank-Lang to write 100 essays on a range of topics and post them automatically to IW. In an extreme sci-fi scenario, a fully language and digitally enabled Frank-Agent bot might be able to live Frank's (digital) life for him.

Even as I was writing this, there is a release of YouAI which promises to learn everything about you and do your bidding via AI, thus digitizing the human mind, which to me brings up all sorts of questions and concerns (to be explored at another time).

Should these tools and scenarios unfold rapidly—only using Frank as an illustration—what does this mean for the human being?

Some questions for thought:

WHAT IS THE MEDIUM/MESSAGE OF IW?

In Understanding Media: The Extensions of Man (1964) author Marshall McLuhan coined the phrase "the medium is the message" to describe the transformational impact of media itself.

From wikipedia:

In Understanding Media, McLuhan describes the "content" of a medium as a juicy piece of meat carried by the burglar to distract the watchdog of the mind. This means that people tend to focus on the obvious, which is the content, to provide us valuable information, but in the process, we largely miss the structural changes in our affairs that are introduced subtly, or over long periods of time. As society's values, norms, and ways of doing things change because of the technology, it is then we realize the social implications of the medium. These range from cultural or religious issues and historical precedents, through interplay with existing conditions, to the secondary or tertiary effects in a cascade of interactions that we are not aware of. [10]

For McLuhan, "the change of scale or pace or pattern" that a new innovation "introduces into human affairs" is the message. Human culture and society changes both due to the medium and the type and speed messages of disseminated through it. There is no clearer example of this than today's social media like Facebook, Twitter, Instagram, YouTube and TikTok. These have altered how information is communicated and shared across many domains of society from medicine to religion to politics.

In contrast to these social media tools, Integral World remains an "old-fashioned" (circa 1990s) digital information board moderated by one generous human individual, Frank Visser who individually edits, posts and manages essay submissions. IW's relatively low-tech, low-bandwidth site affords a certain kind of relationship of IW visitors with its relatively long-form narrative content. One has to go to the site, browse and read the essays, or otherwise click on a link in a notification email. There is no ability for users to submit postings (essays) at will, and there is no algorithmic or AI curation, no personalization and no “essay feed” pushed directly to users through a mobile app. Thus the medium of IW both affords and limits a certain kind of message and message bandwidth through its platform.

I have very much appreciated IW's medium and messages (essays) from a range of writers who thoughtfully reflect on Wilber's and others views related to integral theory, metaphysics, religion, spirituality, science, politics and philosophy. I also consider the “bar” for submitting an IW essay to be significantly higher than posting random thoughts on message forum or discussion thread apps like Facebook or Reddit. I consider the “essay” form akin to constructing a sandcastle to be looked at, and possibly admired, critiqued or even torn down, likely with another essay. For over two decades starting with Frank's first essay in 1996, this has been the medium and method of IW.

Questions for consideration:

ADVOCATING FOR HUMAN ESSAYS

I quickly wrote this short human essay because I feel sad at the potential loss of Integral World as I have known it. It's like when a favorite rock band makes a 90 degree turn and becomes a pop band. There is loss and disappointment coming from the expectation and anticipation of something similarly enjoyable yet also creatively exciting. To change styles or genres can be a let down. The accelerating number of postings is also inhuman, even for a prodigious writer like Frank. I can't digest them. I don't want to digest them. There will likely be five more in the next week. I concur with Elliot Benjamin: [11]

I must candidly convey that I do not think these are “essays.”

I believe that all his other ChatGPT postings should have had a different way for people to get this information—not as Integral World essays; perhaps on a particular place on the site that listed all of Frank's questions to ChatGPT and ChatGPT's answers, or perhaps on a different site.

While I have my own growing pseudo-relationship with ChatGPT, I have a much stronger relationship with IW and its contributors. I enjoy reading many of the essays, not primarily for their content, but for the perspectives each (human) writer brings and shares. I particularly enjoy the comment section because it feels more dynamic and alive with some back-and-forth among contributors and readers.

Like Elliot Benjamin, I want to advocate for original human(s) here on Integral World. I want to advocate for the necessarily incomplete, biased, flawed and impassioned human expression. I advocate for human communication, human debate and human feedback. I advocate for deeper connection between people more than the presentation of ideas. This is my human bias, one that has been growing steadily since Wilber's Integral Theory purported to explain all of human nature and our evolution.

I echo Elliot Benjamin's suggestion that IW separately categorizes AI-assisted postings so that they don't drown out (if “pollute”) human written ones. As a threshold I suggest that if 50% or more of the essay is cut-and-paste or largely inspired or edited by an LLM that it be categorized as an AI-assisted posting. For example, there could be two main categories "Recent Postings" and "Recent AI-Assisted Postings". Or under Recent Postings, there could be latest five human essays (left column) and latest five AI-assisted essays (right column).

FEEDBACK FROM THE IW COMMUNITY?

I encourage comments from IW readers and essay writers: personal, reflective, critical and otherwise. I'm interested in the personal inner experience of what IW and its human (and non-human) essays bring a variety of readers.

For some human reason, I care about IW and its human contributors. I didn't realize how much until the invention of the ChatGPT written/assisted essays on IW. I offer this flawed and incomplete human essay and these suggestions for consideration by Frank Visser and the IW community.

ENDNOTES

[1] Elliot Benjamin, "Frank Visser's Integral World ChatGPT Essays: Are They Essays?", www.integralworld.net

[2] Brent May, "Views from Flatland: What is a Theory?", www.integralworld.net

[3] I asked ChatGPT to summarize my first IW essay:

Q: Read https://www.integralworld.net/may1.html and summarize the main points.

ChatGPT:

[Used WebPilot]

The article titled "What is a Theory?" by B. May on Integral World discusses the nature of scientific theories and their evolution over time, using the example of Albert Einstein's general theory of relativity.

  1. Nature of Scientific Theories: The author explains that scientific theories propose specific causal mechanisms or operating principles and are distinct from everyday hypotheses due to their testability. They are verifiable, falsifiable, and predictive. The author also emphasizes the importance of quantitative observations and rigorous statistical procedures in testing scientific theories.
  2. Einstein's General Theory of Relativity: The article delves into the history and implications of Einstein's general theory of relativity, which unified matter, energy, space, and time. The author discusses how this theory has radically changed our understanding of the universe, leading to concepts like the Big Bang, black holes, and gravitational lenses.
  3. Newtonian Mechanics vs. General Relativity: The author compares Newtonian mechanics and Einstein's general relativity, explaining how the latter reframed the former as an approximation or limiting case. The article also discusses how general relativity has passed all its tests, with visible demonstrations in gravitational lensing, atomic clocks at different elevations or in orbit, and orbital decays of close orbiting binary stars.
  4. Gravity and Space-Time: The author explains how general relativity has shifted our understanding of gravity from a localized object-interactional view to an extensive relational-field view. The article also discusses the implications of this shift for our understanding of the universe's origin and evolution.
  5. Critique of Integral Theory: The author critiques Ken Wilber's Integral Theory, suggesting that it privileges well-delimited structures over extensive contextual aspects, which both transcend and irreducibly impact structure itself. The author argues that modern physics, including general relativity, quantum field theory, and chaos theory, challenges this focus on self-contained structures.
  6. Universe's Expansion and Structure: The article concludes with a discussion on the universe's expansion and structure, explaining how the universe's expansion has led to its cooling and condensation into quantum micro-clumps and gravitational macro-clumps. The author also discusses how structure arises at the localized intersection of the fundamental forces, leading to diverse emergent behavior.

Q: Tell me more about the critique of integral theory in the article

ChatGPT:

In the article, the author critiques Ken Wilber's Integral Theory, arguing that it overemphasizes well-delimited structures at the expense of extensive contextual aspects. The author suggests that this focus on self-contained structures is problematic and not in line with modern physics, including general relativity, quantum field theory, and chaos theory.

The author points out that in modern physics, the concept of a self-contained structure is often replaced by the idea of extensive fields or systems that are fundamentally interconnected and influenced by their context. For example, in general relativity, gravity is not seen as a force acting between two separate objects, but as a curvature of space-time caused by mass and energy. This view sees gravity as an extensive relational field, not a localized object-interaction.

The author argues that Wilber's Integral Theory, with its focus on well-delimited structures, fails to adequately account for these extensive contextual aspects that both transcend and irreducibly impact structure itself. The author suggests that a more comprehensive theory would need to incorporate these extensive aspects to fully capture the complexity and interconnectedness of reality.

In essence, the critique is that Integral Theory, as proposed by Ken Wilber, might be too reductionist or simplistic in its approach, not fully accounting for the complexity and interconnectedness that modern physics suggests is inherent in the universe.

[4] "Stochastic parrot", wikepedia.net

Q: What is a stochastic parrot, particularly in AI?

ChatGPT:

The term "stochastic parrot" is used in the context of AI to describe language models like GPT-3. It was popularized by Timnit Gebru, a former Google AI ethicist, in a research paper that criticized large language models.

A "stochastic parrot" refers to the way these AI models generate text. They don't understand the content they're producing in the way humans do. Instead, they statistically predict the next word in a sentence based on the patterns they've learned from the vast amount of text data they were trained on. Hence, they are "stochastic" (randomly determined) and they "parrot" (mindlessly repeat) the language patterns they've learned.

The term is often used to highlight the limitations of these models. While they can produce impressively human-like text, they don't truly understand the content they're generating, and they can inadvertently perpetuate biases present in their training data.

[5] I asked ChatGPT to analyze this essay:

Q: Looking at the essay "Too Many ChatGPT IW Essays?" analyze the author's affect, assumptions, biases, and potential logical fallacies.

ChatGPT:

Affect: The author's affect throughout the essay is one of concern and frustration. They express a sense of being overwhelmed by the volume of AI-generated essays on Integral World (IW), and a feeling of disappointment that these essays lack the human touch and richness of the traditionally written ones. The author also expresses a sense of loss, as they perceive the shift towards AI-generated content as a departure from the original spirit of IW.

Assumptions: The author assumes that the value of an essay is tied to the human effort involved in its creation, and that AI-generated essays inherently lack the richness and subtext of human-written ones. They also assume that the increase in AI-generated content will lead to a decrease in the quality and meaningfulness of the information available.

Biases: The author shows a bias towards human-written content, valuing it more highly than AI-generated content. They also display a bias against the rapid increase in content production facilitated by AI, viewing it as overwhelming and detrimental to the quality of discourse on IW.

Potential Logical Fallacies:

  1. Slippery Slope: The author suggests that the increase in AI-generated content could lead to a "Boring Apocalypse" where humans are drowned in meaningless information. This is a type of slippery slope fallacy, as it assumes a series of events will inevitably follow from a particular action without providing evidence for this chain of events.
  2. Appeal to Nature: The author implies that human-written essays are inherently superior to AI-generated ones because they are more "natural", i.e., they involve human effort and contain human richness and subtext. This could be seen as an appeal to nature fallacy, which argues that something is good because it is 'natural', or bad because it is 'unnatural'.
  3. Hasty Generalization: The author generalizes their personal experience and preferences to the entire IW community, assuming that others also feel overwhelmed by the volume of AI-generated content and prefer human-written essays. This could be seen as a hasty generalization fallacy, as it draws a conclusion about a large group based on a small sample size.

[6] I tried to get GPT and some plugins to calculate the average per month but it failed. I scanned Frank's essays in the last year or so and 1-4 seems to be the range.

[7] Jonathan Frankle, the chief scientist at MosaicML and a computer scientist at Harvard: 'The "Boring Apocalypse" Of Today's AI', clivethompson.medium.com

[8] "Faulty reward functions in the wild", openai.com

[9] David Christopher Lane and Andrea Diem-Lane, " 'PLEASE DON'T TURN ME OFF!': Alan Turing, Animism, Intentional Stances, and Other Minds", www.youtube.com, 2023.

[10] "The medium is the message", en.wikipedia.org