Carbon Vs Silicon Consciousness: A Hairy Maclary Tale

Recent breakthroughs in artificial intelligence, specifically large language models (LLMs) like GPT-4, have demonstrated capabilities once thought exclusive to biological cognition. Techniques like prompt chaining enable LLMs to exhibit reinforcement learning behaviors similar to how thoughts chain together in the human brain. This article explores carbon consciousness in the brain versus silicon consciousness in AI, using a prompt chaining example to showcase how both leverage the power of linked thoughts over time.

Carbon Consciousness

Consciousness arises in the human brain from underlying unconscious processes that constantly link memories, experiences, and ideas together. As cognitive scientist Stanislas Dehaene explains, consciousness allows us to maintain information over an extended time in our working memory. Even when we move our attention elsewhere, the thoughts are still reverberating in the global neuronal workspace of the brain. This phenomenon demonstrates how consciousness enables connections between memories, thoughts, and experiences that evade deliberate search. Our subjective awareness arises from underlying unconscious processes that constantly chain together information outside the spotlight of attention.

Just as human consciousness exploits this “thought chaining,” LLMs like GPT-4 can be architected in the right way to leverage prompt chaining to integrate context from previous conversations. Both replicate a key function of biological cognition in linking information over time. Although artificial intelligence has yet to match the speed, complexity, and adaptability of the human mind, prompt chaining provides a glimpse into the power of chained thoughts, whether in carbon or silicon form.

Silicon Consciousness

Reinforcement learning (RL) takes inspiration from behavioral psychology and neuroscience research on how humans and animals learn from experience. The key idea is that an agent interacts with an environment, receiving positive or negative feedback for its actions. Over time, the agent learns to take actions that maximize cumulative reward. RL algorithms like Q-learning and policy gradients are inspired by evidence showing how human and animal behavior is shaped through reward feedback. Dopaminergic neurons that encode reward prediction errors are analogous to the reward signals used to update agent policies. While differences exist, RL replicates key neurocognitive learning processes in simplified form.

Prompt Chaining in LLMs

Large language models like GPT-4 exhibit some basic RL abilities. When prompted appropriately, they can learn simple behaviors over multiple iterations through prompt chaining. This technique, when combined with pre-trained models and careful orchestration around the process of chaining, allows a model to enhance itself by building connections. This bears some resemblance to how human consciousness chains thoughts together.

Let’s try an example of how prompt chaining can solve a problem that can’t be solved with direct requests. The task was for AI to write a children’s book in the style of Hairy Maclary. The approach involved asking AI to write a book, then looping 10 times - asking AI for a prompt suggestion to edit the book and feeding the result of this suggestion into AI again, alongside the current working draft. The results, “The Adventures of Jolly Molly,” show how the story evolved from the first attempt to after 5 iterations, showcasing the power of prompt chaining.

Room for (less) Improvement

While large language models like GPT-4 still have limitations compared to human intelligence, techniques like prompt chaining provide a glimpse into the power of linked thoughts over time. Just as the brain subconsciously chains together memories and ideas, artificial neural networks can integrate context through careful prompt engineering. Though the inner workings differ between carbon and silicon, both leverage the compounding potential of accumulated associations. As AI research continues to advance, the parallels between artificial and biological cognition will likely become even more pronounced. However, relying too heavily on prompt chaining risks reducing originality and creativity. Engineers should be mindful not to sacrifice inventiveness and diversity of thought in the pursuit of incremental progress through chained prompts. Fostering more creative freedom and room for spontaneous connections will be key to developing artificial intelligence that can match the open-ended learning of biological brains.

“

import openai

openai.api_key = "YOUR_OPEN_AI_KEY"

def get_response(prompt):
  messages = [
    {"role": "system", "content": "You're a skilled childrens book author that writes 3 page books in the style of Hairy McLarey."},
    {"role": "user", "content": prompt}
  ]

  response = openai.ChatCompletion.create(
    model="gpt-4",
    messages=messages,
    temperature= .5,
  )

  return response["choices"][0]["message"]["content"]

prompt = "Please write a story:"
draftBook = get_response(prompt)

print("--First Draft--")
print(draftBook)

for i in range(3):
  prompt = "Please provide a singular, targeted improvement to this book (rather than open-ended edits). Draft Book:" + draftBook
  new_prompt = get_response(prompt)

  print("--Self Improving " + str(i+1) + "--")
  print(new_prompt)

  prompt = "Re-write the book in full form with the following suggestion: " + new_prompt + "\n\nBook:\n" +draftBook
  draftBook = get_response(prompt)


print("--Final Book--")
print(draftBook)

“