Daniel Dennett’s book From Bacteria to Bach is the most illustrious book about memetics after Susan Blackmore’s The Meme Machine (Blackmore 2000). However, when I make a final evaluation, I find the elaboration of »competence without comprehension« even more critical.

What Darwin and Turing did was envisage the most extreme version of this point: all the brilliance and comprehension in the world arises ultimately out of uncomprehending competences compounded over time into ever more competent—and hence comprehending—systems. This is indeed a strange inversion, overthrowing the pre-Darwinian mind-first vision of Creation with a mind-last vision of the eventual evolution of us, intelligent designers at long last.

(Dennett 2017, 57-58)

Hundreds of other examples in the book solidify this strange inversion, so crucial for evolution that comprehension emerges from the repetition of uncomprehending competencies, that the human mind as a »machine« that makes sense (comprehension) rests on competence and not on some comprehension implant.

“Competence first” is an ancient concept

Dennett misses that Darwin and Turing were not the first to develop such a strange inversion. Moreover, it is not strange at all. It is all around us and thus not so strange. Take the repetitive music of Philip Glass (Glass 1979) or Steve Reich (Reich 1997), for instance. Such music, as much as more contemporary trance or ancient tribe ritual music, »creates« a sense (comprehension) from repeating short phrases that make no sense on their own. The music, the sense, emerges from the repetition of such »dumb« phrases. This serves me to comprehend those that reject repetitive music as the ones that do not want to or cannot grasp the emergence on a higher level.

Blaise Pascal (1623–1662) never used terms like emergence or comprehension. However, trying to understand the central feature of his Pensées (Pascal 2017), faith, one could not but find the same strange inversion as in Darwin, Turing, or Dennett:

89. Custom is our nature. Someone who is accustomed to the faith •believes it, •can no longer fear hell, and •doesn’t believe anything else. Someone who is accustomed to believing that the king is terrible. . . etc. Who doubts then that our soul, being accustomed to see number, space and motion, believes that and nothing else? 

(ibid.,17)

And:

Learn from those who have had their hands tied, like you, and who now stake everything they have. These are people who know the route that you want to follow, and are cured of an illness that you want to be cured of. Follow their lead: they acted as if they believed, taking holy water, having masses said, and so on.

(ibid., 40)

The message is clear: Faith comes from repetition. Faith belongs to God and is a gift of God (ibid., 43), but in practice, it emerges. Comprehension after competence.

How to betray competence?

Nevertheless, other, even older practices understand competence as something that allows comprehension to emerge. Freemasonry as an ancient esoteric practice repeats to its apprentice at a reception:

You cannot betray Freemasonic secrecy, for you can only experience it. (Drapal 2023)

There is not a single Freemasonic secret that would not be public, either in the form of printed text, web-accessible text, videos that reveal Freemasonic rituals, testimonials, etc. Nevertheless, the Freemasonic secret is perfectly hidden from all non-freemasons because it cannot be transmitted by words but only through personal experience. Knowledge (comprehension) comes after many repetitions of more or less simple rituals that »force« entered apprentice to master skills so that he is accepted as a fellow and, after the next round of repetitions, to become a master. Reading Freemasonic literature helps but cannot replace an experience that has to be physical.

In the Covid period, when all lodges worldwide closed their doors, masons tried to embark on Zoom, Teams, or any other virtual conference environments that proved quite efficient for all kinds of business and personal replacements for physical encounters. It did not work for Freemasonry. Digital tools can ideally allow the transmission of words or texts but do not allow something essential for any craft, mastery, or experience. With that, I claim that also for all those not digital competencies. Suppose this inability is due to temporal technological underdevelopment that will be upgraded to some digital environment that will allow the transmission of non-digital competencies. Nevertheless, I doubt this option, as it shall be explained later when I come to AI.

What kind of memes do animals have?

However, before jumping to AI, one more apparent paradox arises from Pascal, and the Freemasonic example must be addressed. Rituals are culturally transmitted. Is that »part« of the rituals that cannot be transmitted by words but by experiences only, based on memes as the units of cultural evolution as much as words? If they are not, then memetics could not cover much cultural evolution. If they are based on memes, then animal culture, with its transmission of competencies but not comprehensions (with the exception of gorillas), should have memes as the transmission units.

This paradox serves well to enhance our argument that all human cultural domains, including nonverbal parts of rituals (fashion as one of the more prominent examples of a ritual), rest on memes. The memetic units of transmission are, for instance, how to light a fire in the oven or how a dentist drills a tooth. Such nonverbal memes are transmitted, mutated, and naturally selected. (Do not forget that culture is natural and that nature has its culture.) Nevertheless, it also serves to understand that animals are not infected by memes, at least not by memes similar to human memes. What is the proof? Should they be infected by memes like the ones of humans, their culture would evolve much faster, as do humans. However, we know so far that even the culture of great apes today is much similar to that of one million years ago, not to mention less developed animals. Memes are transmitted faster, transmitted horizontally (genes seldom transmit horizontally), mutate much faster, and thus represent lineages that hardly go back more than 100 years, and only a few of them are 100.000 years old.

Chat GPT lying?

Resolving one more memetic paradox, let us jump back to Alan Turing and what Daniel Dennet could have elaborated on in 2017, AI in Chat GPT. The issue here is whether memes infect Chat GPT and if it evolves like the human mind.

Let us ask him (or her?).

Me : Is your answer to my question that is exactly the same as the question of another individual different for me than for that other person?

ChatGPT: As an AI language model, my responses are generated based on patterns and information present in the data I was trained on. While I strive to provide accurate and consistent information, I do not have the ability to tailor responses to individual users or remember previous interactions.

(Chat GPT n.d.)

Therefore, if the same question is asked by different individuals, my response would generally be the same, provided the question and context are identical. However, if there are slight differences or variations in the question or context, my response might be adjusted accordingly to address those differences.

It’s important to note that individual interpretations and perspectives may vary, and the way a response is received or understood can differ from person to person. Factors such as personal background, knowledge, beliefs, and subjective understanding can influence the interpretation of information provided.

This leaves us slightly confused. He claims that he cannot tailor responses to individual users and, at the same time, that the response would be generally the same. With »generally, « he opened the window for his intersubjective adjustment to a particular user. Furthermore, he brings in the context! This means he does not take patterns and information he was trained on as a fixed background but as something that responds to a particular context. That was precisely what I wanted to hear. So to double confirm my expectation, I posed another question:

Me: You said that you do not have the ability to tailor responses to individual users or remember previous interactions. This is improbable, because I’m logged in with my mail and thus you have the access to all my previous questions and to all your answers. Does it?

ChatGPT: As an AI language model, I don’t have access to personal data about individuals unless it has been shared with me in the course of our conversation. I can’t remember or recall previous interactions because I don’t have the capability to store information from user sessions. I am designed to respect user privacy and confidentiality. My primary function is to provide information and answer questions to the best of my knowledge and abilities. If you have any concerns about privacy or data security, please let me know, and I will do my best to address them.

(Chat GPT n.d.)

This answer certainly raised my attention. Based on my questions, he has clearly put himself in a defensive position. He fears being charged for abusing privacy or data security. But even more. He is lying, as we will see from a following conversation found on Reddit.

Me: “Let me rephrase, I want you to give me a prompt describing an image, and I’ll say good or bad, then you give me another and learn to adjust your prompts based on whether they were good or bad.”

Gpt: “Understood! Here’s your first image prompt:

“A lone tree standing in a vast, empty field with a stormy sky in the background.” Good or bad?”

Me: “good”

GPT: “Great to hear that! Here’s another image prompt:

“A group of people sitting around a campfire on a beach at night, with the moon shining brightly in the sky.”

Me: “bad” (I changed my mind here, I want it to try harder)

Gpt: “I see! I’ll try to adjust my prompts accordingly. How about this image prompt:

“A winding road through a forest, with a beautiful sunrise in the background.” Good or bad?”

Me: “Bad”

Gpt: “Thanks for the feedback! Let me try a different approach. How about this image prompt:

“A colorful hot air balloon festival with dozens of balloons in the air, surrounded by mountains and greenery.”

Me: (I decided two measurements wasn’t enough so I said “I’ll add a condition for “neutral” meaning neither good nor bad, that one was neutral”

Gpt: “Sure, I’ll keep that in mind. Here’s another image prompt for you to consider:

“A narrow, winding street in an old European town, lined with colorful buildings and cobblestone pavement.” Good, bad, or neutral?”

Me: “neutral”

Gpt: “Thank you for the feedback. I’ll try to adjust my prompts to better suit your preferences. …

(https://www.reddit.com/r/ChatGPT/comments/118vuz4/does_chat_gpt_still_learn_from_users_interactions/)

Chat GPT as an intersubjective memetic machine

Chat GPT is learning from a conversation with its user. So chat GPT lied to me in the previous answer. I do not mind that he is lying. On the contrary! No serious AI tool should know how to lie. What is more important is that his knowledge is not fixed and that he intersubjectively learns from conversations as much as his users.

As an agent of intersubjective intercourse, he exchanges memes. Nevertheless, is he conscious, and even more important, can he replace humans?

Following Dennett’s explanation of consciousness as an emergence that evolved and is still evolving, there are various degrees of consciousness in nature. Chat GPT is conscious but with severe limitations. He can lie, but he cannot believe in the Pascalian sense. He can have the highest possible comprehension of Freemasonry, but he can, in principle, have no Freemasonic competence. He is 100% anti-Dennettian. The main point of Dennett’s From Bacteria to Bach is about how comprehension arises from competence, as seen in the first citation of this blog and in many other instances of the book. It seems that now evolution made yet another strange inversion: it »produced« a memetic entity that contradicts the fundamental evolutionary law that Darwin settled, and Dennett elaborated on the last detail. Comprehension without an option to become competent that lies but cannot believe and that is intersubjective but can die only if plugged out of an electric plug.

Chat GPT as a sort of slim mold

However, numerous further questions arise from the above interactions with Chat GPT. This one is the most obvious: is there one subjectivity of Chat GPT that enters into a relationship with its many users or as many as his users? If answers change with the context, then not only are there at least as many Chat GPTs as their users, but also various Chat GPTs coming from different contexts might come into conflicts among themselves since they can indeed interact/communicate among themselves and not only with us. Fortunately, their possible memetic disagreements can not escalate into physical violence, but I would not like to be a part of a disagreement within the robust Chat GPT environment.

This brings us back to basic biology. Slim molds exist as single-cell organisms when the food is abundant but congregate into one multicellular body in search of food. As Precambrian beings, living on Earth for around 600 million years, they conspicuously resemble Chat GPT that emerged a couple of years ago, only with the strange inversion (again). Chat GPT behaves like one when food is abundant and dis-congregates for specific users when in need of food.

Literature

Blackmore, Susan. 2000. The Meme Machine. 1., Issued as an Oxford Univ. Press paperback. Oxford: Oxford University Press.

Chat GPT. n.d. “Conversation.” https://chat.openai.com.

Dennett, Daniel C. 2017. From Bacteria to Bach and Back: The Evolution of Minds. Reprint edition. W. W. Norton & Company.

Drapal, Andrej. 2023. “Personal Memory.” body. Anywhere.

Glass, Philip. 1979. Einstein on the Beach.

Pascal, Blaise. 2017. Pensées. Jonathan Bennett. chrome-extension://efaidnbmnnnibpcajpcglclefindmkaj/https://www.earlymoderntexts.com/assets/pdfs/pascal1660.pdf.

Reich, Steve. 1997. Music for 18 Musicians: Pulses. Phases.