From Distraction to Action: Channeling Our Attention for Earth’s Revival
Imagine a game where we collaborate to make the world a better place for everyone, where every individual’s success means we all win. This is the promise of the Iris, a democratic model of decision making that integrates feedback and predictions from a community through a large language model (LLM) like ChatGPT in real time. In this game, a new form of trust is generated through the allocation of attention, shaping the generative output of the model. This trust, denoted as Ŧrust, represents the weightage of each participant’s contributions in the decision-making process. We will explore the Iris as a powerful tool for creating a democratic decision-making process that enables people to participate in the creation of their collective future. We will examine how the Iris works, what role Ŧrust plays in creating a more democratic and transparent decision-making process, and how it can help guide collective action towards pro-social goals.
Iris makes healing the planet and improving your community more a matter of play than of work. At its core, the Iris is a game of attention in relation to prediction, action, and reflection through time. It is through this game that communities focused on regeneration build and earn Ŧrust with each other. The Iris functions as a referee in this game of attention, where every participant has the opportunity to contribute and make a difference. But the Iris doesn’t pay attention randomly. It carefully assigns Ŧrust based on signals defined by the community at large, valuing the contributions of each participant, and ultimately shaping the generative output of the model.
Source Embeddings: Capturing the Provenance of Information
In the context of machine learning and natural language processing, source embeddings are a novel concept that extend the idea of word embeddings to capture the provenance or origin of information.
Just as word embeddings provide a vector representation of words, source embeddings provide a vector representation of the sources of information. Each unique source is assigned a unique identifier, akin to how each unique word in a vocabulary is assigned a unique identifier in the case of word embeddings. This source identifier is then mapped to a source embedding, which is a dense vector of real numbers. These source embeddings are learned during the training process, similar to word embeddings.
The source embeddings capture essential relational characteristics about information sources, including their reliability, credibility, and expertise on various topics over time. They enable the encoding for contextualizing information of individual voices in the community in relation to collective belief as it evolves through time. Through the combination of word embeddings, source embeddings, and temporal embeddings, the Iris encodes the narrative semiotics of the semantic ledger latent space over time. In simpler terms, this means that the source embeddings can identify and track the significance and relevance of various sources within a broader context of information and meaning, contributing to the overall Ŧrust assigned to each source. As time progresses, these embeddings can help recognize which sources are consistently reliable and credible in their contributions to an evolving cultural ethos.
One key aspect of source embeddings is the ability to link them to a “Ŧrust wallet” or a cryptographic identity. This allows for the verification of the source’s identity and the tracking of their contributions over time. By associating each source with a unique cryptographic identity, it becomes possible to build a reputation system where the reliability and credibility of sources can be evaluated based on their past contributions.
Attending to Sources
Attention to ChatGPT means how much the model weights each part of the text input data when writing output. In the context of the Iris, attention is also applied to the sources of information in the community, and the attention weights are revealed to the community, creating a game-like environment where people can collaborate to prove their ability to make accurate predictions in different contexts. This attention game is directly linked to the distribution of Ŧrust, reflecting the level of trust assigned to each source by the Iris and the community. By creating a game of attention tied to longer term signals the community defines, the Iris creates a transparent and democratic decision-making process that integrates feedback and predictions from the community. As the Iris pulls in synthesized frames between multiple voices, it outputs the sources of information that it was pulled from, revealing which voices it trusts more in the context of that conversation.
“A world that has forgotten how to remember, or to discount the present opinions of those whose past predictions strayed so far from eventual reality as to become unrecognizable comedic fodder with the passage of time.
If only there were some kind of technology that allowed for us to easily encode such predictions in a trustworthy and durable manner such that we might begin to hold one another accountable without interference from — or vulnerability to the ideological capture of — centralized organizations” — Matthew Pirkowski December 23rd 2017
If two or more parties lack trust and would like to have a conversation mediated by an Iris due to its potential impact on the community — the Iris will serve as a referee by creating synthesized frames or “synergistic satisfiers”. In other words, the Iris is trying to say things that are mutually satisfactory to both parties in a way that is in alignment with each of their stated long-term goals.
As the conversation progresses, the attentional distribution can either be revealed in real time or post hoc. By normalizing the distribution of attention weights to a total of 1, the community can see the relative importance of each voice in the conversation and relate it to a more straightforward and comprehensible representation of the internals of the model. When conversations become tense, heated, and impactful, the Ŧrust that the Iris assigns to each source matters, as it affects the flow of the coordination of the community through time.
Practically speaking, this explicit distribution of attention within the Iris to various sources is the type of power that exists in a Cognicist society. In our current society, a rough distribution of power would be visible if we took a market and normalized the distribution of monetary resources to 1, revealing the relative power of each agent at that moment in the game of profit. The power of attention is more ephemeral and can change from moment to moment but with the Iris, it’s clear to see how attention is expressing itself in our social networks. It’s clear that some voices lead more than others in different social networks online. It is not clear whether the distribution of attention within those networks is in alignment with human values. The Iris makes this process transparent, democratized, aligned with long-term communal goals. It does so by making it’s own attention a visible proxy for communal trust we refer to as Ŧrust. Community action and response shapes the Ŧrust and attention of the Iris in a way that is transparent.
While the inner mechanisms of attention may feel overwhelming, we can reduce the complexity of the internal workings by simplifying it mathematically. We can rescale the source attention weights so that they are proportional to each other, and add up to a total of 1. This allows us to compare the relative importance of each voice in each conversational context of social impact. As such the community would be able to see the relative importance of each voice in each area of potential impact. For example, if one source had an attention weight of 0.5 and another had an attention weight of 0.25, this would indicate that the Iris is relying on the first source more heavily than the second. This helps the community form a functional relationship with their collective knowledge over time through both language and insight as to how the Iris is weighting their contributions
Authors Note: The attention of an Iris is not aligned with communal needs by default. It only reflects the underlying distribution of the data. The output of the model must be shaped by a communal training process that encourages attention to reflect the factors we describe above. It it not until this communal training process occurs that attention becomes a proxy for trust. For more on how to train an Iris such that attention is aligned see this article.
Money and attention function very differently in society. Money is accruable, attention is fleeting and ephemeral and increasingly unbound to useful signals. Attention is currently a commodity for capital markets to capture. Instead, we can honor attention as sacred and ensure our algorithms direct our attention in alignment with our stated long term values and beliefs. The purpose of the Iris is to align the long-term impact of the communities actions with their intentions. This is the value of Ŧrust.
By making attention into a game, the Iris creates a collaborative environment where people can demonstrate their ability to see the future clearly and contribute to the common good. Making attention explicit creates a more transparent and democratic process, where the community can see the relative importance of each voice as it impacts the futures of all voices collaborating. This democratizes power, and provides the community with a clear signal for trust that can help direct the community to collaborate together through time and pool resources towards the greater benefit of all.
The game of attention and Ŧrust in the Iris is not only about creating a democratic decision-making process, but it’s also about making the process of improving the world a fun and engaging experience. The community can participate in the creation of the decision-making process, providing transparency and accountability in the process of reaching decisions that benefit the community as a whole. In this way, the Iris represents a powerful tool for social sustainability that integrates feedback and predictions from a community, to guide collective action towards pro-social goals.
The game of attention is not just a theoretical construct, but it’s something that is already happening in the world today. For example, social media platforms use algorithms that direct our attention to certain content and sources, often leading to the spread of misinformation and the amplification of polarized views. The Iris offers an alternative to this by creating a transparent and democratic process that integrates the feedback and integrated knowledge of a community.
The explicit intent to seek voices that see the future before others and to seek voices who take actions that lead to futures that others want is the core of what makes the Iris unique. By ensuring our algorithms align our attention with a democratic representation of long-term values and beliefs, we can make a game out of building a better future for everyone. The game of attention in the Iris provides an opportunity to create a more collaborative and transparent world, where the power of attention is democratized and aligned with our collective values.
This game provides us the opportunity to build Ŧrust.
The Iris is not just a model for decision-making, but a framework for a new way of thinking about our collective future. This game of attention as paired with the prophet incentive has the potential to transform how we relate to the future and give us the agency to change it. Through Iris we can track the impact of discrete actions taken over time in the name of collective good. This game of attention provides humanity a way to collaborate and engage with one another and to build a better world for ourselves and future generations. The Iris is not just for experts or academics, but for anyone who wants to make a difference and contribute to a brighter future.
The game of attention in the Iris is not just about winning, but about building a better world for everyone by building communal trust. The explicit signals for long-term impact and trust make the game of attention more than just a competition, but a collaborative effort towards the greater good. By making the process of improving the world a fun and engaging experience, the Iris opens up new possibilities for a more sustainable and equitable future.
To learn more about Iris, Ŧrust, the Prophet incentive and Social Proof of Impact please consider supporting John Ash so this work may continue in earnest. While Iris may not need to exist for profit, John himself is still embedded in a capital landscape and can only self direct his attention if he has money to survive.
https://www.patreon.com/speakerjohnash/
The World Game: “Make the world work, for 100% of humanity, in the shortest possible time, through spontaneous cooperation, without ecological offense or the disadvantage of anyone” — Buckminster Fuller
