Can AI & Advanced Technology Address ESG Challenges?
Part 1 of a ChatGPT deep-dive, providing much-needed context for a disruptive technology
OpenAI's impressive(?) ChatGPT — an AI-powered chatbot and its ability to interact with humans conversationally — has taken the world by storm. It achieved 1m users in less than 5 days.
When a new technology makes such an impact, it’s worth paying attention to.
In this week’s newsletter, I’ll summarize the fuss and then make suggestions on how a technology like ChatGPT could be used (but isn’t currently) to address the challenges that ESG purportedly addresses.
I should let you know that parts of this newsletter are written by ChatGPT. I will clearly indicate the sections that are, including what prompt I used to retrieve the text. By showing actual examples, you’ll better get an idea of where this tech could go.
What’s in this Post?
ChatGPT Deep-Dive (7min read)
Vertical Learning
ChatGPTs Many Biases
The Sahtouris Keyboard Metaphor (2min read)
A ChatGPT Thought Experiment (2min read)
What I’ve Been Reading & Watching (links to +/- 50min reads)
ChatGPT Deep Dive
The interwebs are already abuzz with a ton of content explaining what ChatGPT does and what it means for humanity. Rather than repeat what can already be found online, I’ll explain why I’m disappointed by the use cases proposed and where a more exciting exploration might lie.1
First, I’ve done a ton of homework on ChatGPT. I’ve run over 100 iterative experiments (will explain what this means shortly), watched numerous interviews of the big names behind the tech and I’ve bought about a dozen guides, all of which promise the earth and deliver very little. The guides breathlessly expound on what ChatGPT means for the future of coding, blogging, poetry, music, art, education, and even book authoring. The interviews are mostly philosophical, with many debates about the future of work, existential questions about the meaning of life, as well as concerns about copyright, intellectual property and what creativity is (art produced by a human vs. art produced by a machine) and how we incentivize creativity. In short, ChatGPT has created quite a stir.
Sadly, there’s very little new ideation. Mostly it’s all about how we continue doing what we’re already doing, just faster and with a few shiny bits added in. In fact, most of what I’ve seen reminds me of the very first websites: clunky and not particularly inspiring:
What we know from history, though, is that early technologies improve very quickly, and there’s no reason ChatGPT won’t do the same.
That said, ChatGPT — even as it stands today — is pretty impressive, depending on how the technology is used. Basically, the quality of ideas and the understanding available to us depends on the quality of the prompts (the natural language input a user types into ChatGPT to generate a response). I’ve seen many examples where an existential question is keyed into the prompt, with fairly standard Wikipedia-type outputs. (That’s a low-quality prompt.)
If you’d like to try this experiment yourself, enter the following prompts:
Provide a list of existential questions
Expand on point 1 (only once the list to the first question has been provided)
What does this mean for humanity? (once the above question has been answered)
Expand on point 2
If you repeat this process until you get bored, you will notice that the machine starts referencing the information it previously provided. The answers are slightly richer than if you simply enter one of the existential questions (as a low-quality prompt) without setting the context first. This is what I meant by iterative experiments, which leads us to Vertical Learning.
Vertical Learning
Before explaining Vertical Learning, it’s worth reminding ourselves of one of the big (but simple) breakthroughs the internet provided: hyperlinks.
Here’s a ChatGPT response about hyperlinks after I asked the following prompts in sequence:
How did the introduction of the internet impact human development?
What's the significance of hyperlinks within this transformative impact on human development?
Summarize your answer.
The response:
ChatGPT: Hyperlinks are an essential part of the internet, and they have had a significant impact on human development by making it easier for people to access information and knowledge, communicate with others, and stay connected. Hyperlinks have contributed to the democratization of knowledge and have transformed the way we learn and interact with each other.
This is a very different blah-blah-blah response to the low-quality prompt of “What is a hyperlink?”:
ChatGPT: A hyperlink, or just link, is a reference to a web page or other resource that you can click on to access. It is usually displayed as blue, underlined text, and clicking on it takes you to the resource that it refers to. Links are an essential part of the web, as they allow users to navigate from one page to another by clicking on links that contain the URL of the destination page. Links can be found in various places on a webpage, such as the body of an article or in the navigation menu, and they can also be shared in emails, social media posts, and other online communications.
So, stacking ideas, concepts or contexts one on top of the other provides a richer ChatGPT experience than when asking a simple question. This is essentially Vertical Learning.
There is no succinct definition of Vertical Learning online.2 Here's my attempt:
Vertical learning leverages current understanding to build new understanding. It is the ability to hold more and more complexity, to hold increasingly more perspective in one’s awareness. Where Horizontal Learning is composed of the individual bricks of knowledge, Vertical Learning is how the house is constructed using those bricks.
“…vertical learning not only upgrades a leader’s operating system, but it also rewires their hardware and expands their leadership competency and capacity. It literally alters brain functioning and recreates a leader’s worldview.” - Barrett C. Brown, PhD. The Future of Leadership for Conscious Capitalism
This idea of “recreating a worldview” is what excites me about the possibility of ChatGPT. However, we will never accomplish this if we only ever approach “the machine” within our current worldview.
The Many Biases of ChatGPT
Identifying the inherent bias in OpenAIs tools is where ChatGPT is particularly useless. The best I can get out of it is the following response to this prompt: “How are large language models biased?”
ChatGPT: Language models are trained on large amounts of text data, and the data they are trained on can contain biases that are present in the real world. For example, if a language model is trained on a dataset that contains a disproportionate number of examples written by a particular demographic group, it may learn to generate text that is more similar to the writing style of that group and may exhibit biases present in the training data. Similarly, if the training data contains stereotypes or other biased language, the model may produce biased output.
Asking specific bias questions about OpenAI or ChatGPT or even the datasets used to train ChatGPT results in a huge and frustrating waste of time.
What we do know is that OpenAI is founded by big Silicon Valley names (and egos): Carlos Virella, Elon Musk, Greg Brockman, Ilya Sutskever, James Greene, Sam Altman, Wojciech Zaremba. The current CEO, Sam Altman, is a 2016 WEF Young Global Leader graduate, along with Emmanuel Macron and Lila Tretikov (now a Microsoft AI vice president).3
So without doing too much digging, we know that ChatGPT has capitalist, colonialist, and neoliberal biases (an answer you cannot get from ChatGPT). If we take into account the lack of indigenous wisdom included in its training data (because indigenous wisdom is orally recorded not text-based), we could go as far as saying that ChatGPT is itself infected with Wetiko, the cannibalistic spirit that embodies greed and excess.4
This inherent bias leads us on to a discussion of where ChatGPT sits on the scale of human consciousness, rather than where it sits on the scale of technological accomplishment.
The Sahtouris Keyboard Metaphor
Elisabet Sahtouris is an evolution biologist with a deep interest in Eastern and Western scientific worldviews. I particularly enjoy her reliance on metaphors to explain complex topics:
“The task of science is choosing the appropriate metaphors because metaphor is the only way to convey the scientific mode.” — Elisabet Sahtouris
One of my favorite metaphors of hers is the Keyboard Metaphor, which she has been talking about for many years in conferences, interviews and podcasts. The Paradigm Explorer magazine finally published a 1-pager of her metaphor in February 2022.5 As far as I know, this is the only written version and it's brilliant. It describes matter as existing in the low notes of the keyboard and spirit as existing in the high notes. Western science attempts to make music using only the low notes. Eastern science makes music by including the mysticism covered by high notes. Humanity’s next great symphony will be written when we can play with all the notes, high and low plus everything in-between, as one infinite keyboard.
ChatGPT, and all Large Language Models, exist within and are constrained by the low notes of this keyboard. So are all the technocratic ‘solutions’ proposed by the WEF and their ilk. So is ESG itself. Every single requirement to measure and report on matter — whether that’s carbon emissions or inequality — keeps us trapped in the low notes.
So, can ChatGPT address ESG challenges?
When we revisit Barrett Brown’s definition of Vertical Learning, we notice that what is required is a literal alteration of brain functioning and recreation of our leaders’ worldviews.
ESG does not accomplish this shift in worldview. Nor does ChatGPT.
But could the merging of human genius and creativity combined with the computational power of Large Language Models assist in stacking known ideas, concepts or contexts one on top of the other to “recreate a worldview” that actually works? (There are many already-existing alternative worldviews. Sadly many of them have little hope of moving from the theoretical to the “in-widespread-use” stage.)
A ChatGPT Thought Experiment
In a future newsletter I’m going to compare the leap forward these technologies have and could produce:
Hyperlinks and their ability to quickly link information (Horizontal Learning)
ChatGPTs potential to link disparate bodies of knowledge to extrapolate previously unexplored Vertical Learning by tapping into Collective Consciousness.
To explain, here’s a thought experiment:
One hundred individuals, all highly accomplished in their field of expertise, come together;
Over a period of time spent building trust with each other, these individuals share 6-10 trials and 6-10 accomplishments from their lifetime amongst themselves;
Through a structured process — perhaps aided by ChatGPT or other Large Language Models — patterns between these moments are explored;
Perhaps these patterns can be constructively shaped into a narrative:
One of the iterative experiments I ran was to feed highlights and lowlights of my life into ChatGPT, asking it to write A Hero’s Journey outline and then the story itself, using my inputs. The result, while not factually accurate, was pretty emotionally moving.
Another iterative experiment I ran was to ask for possible narrative patterns that can be used as a template for self-reflection and introspection to discover one’s life purpose. After a few false starts, it suggested The Four Noble Truths, The Five Stages of Grief, The Wheel of Life, and The Eightfold Path. It then suggested open-ended questions for each of the narratives.
The point is that from a ‘high note’ perspective of the keyboard, each of these individuals has a higher purpose for their lives. In the hustle and bustle of their ‘low note’ lives, they seldom explore what this purpose might be, except as part of a very ‘high note’ meditation exercise, or plant-inspired journey. They never collectively explore what the purpose of our species might be, taking Collective Intelligence and the human experience into account. What ChatGPT has (somewhat) accomplished is to collate the human experience into a text database and make it widely available for questioning, in a natural way. The way it’s being used now is for very limited creativity within given mental models. We’re still playing in the same playground. The real opportunity is to build on these mental models to compose entirely new constructs of how the world works; to create entirely new playgrounds of governance and incentive systems.
Can ChatGPT and other AI tools — when merged with deep human creativity and experience — help us accomplish reaching into the collective intelligence that already wants to play in a new playground? The playground equipment already exists in the form of decentralized governance and decision-making; citizen-issued complementary currencies; bioregional food and energy production; and more. If we can explore the use of these tools along these lines, there may be hope. Otherwise, we’re just repeating past mistakes.
That’s all for this week, but I sense the exploration still has a long way to go.
Michael
PS. If you’re aware of active collective intelligence explorations already underway, please do let me know. I know of many theoretical experiments but haven’t seen anything concrete yet. If this post triggered any creative ideas, please go out and build on them!
PPS. In next week’s newsletter, I’m going to explore the dark side of ChatGPT. Specifically, how does ChatGPT fit into Frederick Hayek’s notion of the Use of Knowledge in Society, as he referred to it in 1945?
What I’ve Been Reading/Watching This Past Week
My focus over the past week has been on social engineering as a top-down psychological manipulation to influence particular attitudes and social behaviors on a large scale in order to produce desired characteristics in a target population. That’s what I sense is going on with ESG, so I’ve explored the ‘dark side’ of social engineering.
https://en.wikipedia.org/wiki/Social_engineering_(political_science)
https://www.quora.com/What-is-social-engineering-in-politics
https://social-epistemology.com/2018/09/18/humanitys-halting-problem-adam-riggio/
If you’re unfamiliar with ChatGPT and want a quick introduction, head on over to our research portal, where there is enough of an intro, without getting you bogged down in too much of the overwhelming detail: https://bit.ly/chat-GPT
For the history of Vertical Learning, see https://bit.ly/Vert-Learn
Note that it’s now almost impossible to find meaningful results for YGL graduates using the usual search engines. Fortunately, https://maloneinstitute.org/ maintains a regularly updated list, available freely for download.
For a deep-dive into Wetiko, see https://bit.ly/Wetik0