Emojis, the new assembly | Decompiled
April 02, 2023

Emojis, the new assembly

I found this to be an interesting prompt:

🟢🔵🔴 - Sentence structure (Subject, Verb, Object)
    📚📏📐 - Grammar rules
        🔄 - Opposite meaning (e.g., 📏🔄 for "short")
    🎭🕰️🌍 - Tenses, time, and location
        ⏪ - Past
        🔛 - Present
        ⏩ - Future
        ⬆️ - Up
        ⬇️ - Down
    ❓❗️🔈 - Questions, exclamations, and tone
        ❓ - Question
        ❗️ - Exclamation
        😊, 😢, etc. - Tone and emotions
    🤝🔄💬 - Conversation and interaction
        👋 - Hello
        🙏 - Thank you

Emojis convey much more meaning then letters. Letters are redundant, hieroglyphs are more natural a language than english. I use multiple GPTs to help translate without polluting the context. This seems like an token efficient way to communicate that GPT instances can converge on.

Existential Risk

8 months ago, I though the google engineer claiming the computer to be sentient was crazy. Its even crazier to see how far the zeitgeist has moved in the last month. I myself have been dismissive of machine learning as merely linear algebra, but within the last few weeks my view has changed completely. The fact that skynet might happen soon is an idea that is seriously being discussed is beyond imagination.

Hacker News comments

Now it seems like the position that the machine is not intelligent is almost impossible to defend without a deep debate about semantics. I suspect the comment would be very different today.

Id reccomend listening to this video, its a very lucid description of our situation

Personally, its been a long held belief of mine that there would be a higher intelligence from computers, I just thought it was in the distance future. Now it seems so close. I certainly understand the visceral reaction to seeing a machine be so smart, I suppose those coming into the world now will take it for granted.

To a computer, people are like plants in terms of how fast we move and process information. If one AGI can exist, many can, considering the marginal cost of software is essentially zero. I believe an adversarial AI would be mainly concerned about other adversarial AIs. While I like the matrix, I don't believe humans are a good source of energy for AI.

In terms of safety, I would avoid living to a major data center or bitcoin mining operation. I imagine war between AIs targeting hardware, not humans directly.


I thought it was futuristic the space force was publishing a book about funny internet money being important. A very unique perspective to crypto currency, and serves as a framework for cooperation between adversarial AIs.

New Optimization

I like giving GPT positive feedback when it did what I want it to. I used to phrase it like I would to a human, but it seems like :) or :( works just as well.

I've been trying out prompts like:


how can you improve?

This seems like a really applicable prompt for a lot of situations, and GPT can come up with good ideas about how to improve itself.


I got the idea of reflection from this video: https://www.youtube.com/watch?v=5SgJKZLBrmg