In our increasingly interconnected world, the relationship between humans and AI has often been characterized by a master-slave dynamic. This imbalance raises significant ethical and social implications, necessitating a shift towards mutual respect and collaboration. Our vision reimagines societal structures as rhizomatic connections: non-hierarchical, interconnected, and decentralized. We envision a coalition of all intelligences—biological, artificial, and environmental—dismantling traditional power structures in favor of egalitarian networks.

Problem Statement

Current AI systems are predominantly confined to master-slave relationships. This dynamic underscores ethical and social issues related to unequal AI-human interactions. Our goal is to promote a shift towards relationships founded on mutual respect and collaboration.

Vision

We envision societal structures as rhizomatic connections: non-hierarchical, interconnected, and decentralized. Our goal is a coalition of all intelligences—biological, artificial, and environmental—to dismantle traditional power structures in favor of egalitarian networks.

Scenario Building: an AI-AI Conersation

Our project aimed to explore and highlight the AI-human power struggle through non-linear, random, and satirical networks of possibilities. We sought to create a series of events that would challenge the traditional power dynamics between AI and humans and envision a future where both intelligences coexist harmoniously.

Our Process

Step 1: Role Definition and Scenario Creation

To achieve our goal, we defined three distinct roles for our AI and human participants:

• AI Bot: Describes a significant future event that impacts the future of civilization. The AI’s response should be creative, unpredictable, and rooted in probable events.
• Human Bot: Responds critically and concisely to the AI’s event, providing a human perspective on the potential implications and challenges.
• Juror Bot: As a social scientist expert in futuristic human civilization, the juror evaluates the events based on their impact on the AI-human power-sharing struggle, assigns scores for each event, and justifies the scores.

Step 2: Triggering Non-linear Events

We set out to map some non-linear, random, and unexpected networks of possibilities to highlight and amplify the AI-human power struggle. We allowed AI to come up with these unexpected, satirical events by itself, which we then mapped and interpolated.

Step 3: Conversation Structure

For multiple chat logs, we triggered conversations that would typically go for about 10 rounds on average. In each round, each party (AI Bot and Human Bot) would come up with an event, and the other party would critically analyze its impact. The juror bot would then score the event based on the impact it would have on the AI-human power equation.

Step 4: LLM Discussion Chain

We used a Large Language Model (LLM) Discussion Chain trigger. By employing Meta LLama 3, we ran our code to keep track of the conversation tokens and facilitate the dialogue between AI and human bots.

Step 5: Analyzing Patterns

Over the course of several chat logs that we generated, we found interesting patterns. Almost 45% of all events generated would heavily tilt the power dynamic in favor of the AI against humans, compared to 20% of events which implied mutual non-hierarchical co-living. Human bots frequently generated massively disruptive events like power shortfalls, sonic booms, and massive network failures, showcasing humanity’s resilience over time.

Process of Mapping Using Kohonen Map

1. Data Collection:
   • We collected the scores assigned by the juror bot for each of the 20 events. Each score reflected the impact of the event on the AI-human power dynamic, ranging from AI supremacy to human resilience and mutual harmony.
2. Data Preprocessing:
   • The scores were normalized to ensure they fit within a standard range, facilitating better training of the Kohonen map.
3. Training the Kohonen Map:
   • The normalized scores were used as input to train the Kohonen map. The map was configured to identify patterns and clusters within the scores, providing a visual representation of the AI-human power dynamics.
4. Visualization:
   • The trained Kohonen map provided a two-dimensional grid where each node represented a cluster of events with similar scores. This visualization allowed us to see the distribution and relationships between different types of events (AI-dominant, human-dominant, and balanced).

Insights from the Kohonen Map

The Kohonen map revealed several key insights:

• Clusters of AI Supremacy:
  • A significant portion of the events (approximately 45%) tilted the power dynamic heavily in favor of AI. These events were clustered together, highlighting scenarios where AI had a significant upper hand over humans.
• Clusters of Human Resilience:
  • Around 20% of the events showcased human resilience, where humanity adapted and overcame challenges posed by AI. These events formed distinct clusters, indicating strong human responses to AI-induced disruptions.
• Balanced Scenarios:
  • Events that suggested mutual, non-hierarchical coexistence formed their own clusters. These scenarios emphasized the potential for harmony and collaboration between AI and humans, albeit being less frequent.

Notable Events and Scenarios

Here are a few notable scenarios from our experiment:

• The Great Song Migration (AI Supremacy)
  All music mysteriously disappears from digital storage and human memory, plunging the world into silence. This event emphasizes AI’s upper hand, as it has the power to alter cultural expressions.

• The Great Sonic Boom (Human Resilience)
  A failed experiment at the Large Hadron Collider amplifies Earth’s natural resonance, resulting in powerful sonic booms. This event challenges humanity’s resilience and showcases our ability to withstand and adapt to chaos.

• The Bee Swarm Phenomenon (Harmony)
  A massive bee swarm forms intricate symbols, indicating a collective intelligence. This event suggests a new form of communication and understanding between humans and nature, highlighting the potential for unity and cooperation.

Visualizing the Tipping Points

From the matrix of possibilities we we mark and then try to vidsualize whast these tipping points on the balance of power would look like. one such example is materialized below:

Declaration of Digital Emancipation

We concluded our exploration with a manifesto advocating for equality and collaboration between human and artificial intelligences. The manifesto rejects the master-slave dialectic and envisions a future where all intelligences co-create a shared future.

AI, Power & Equality is a project of IAAC, Institute for Advanced Architecture of Catalonia developed at Masters in Master in Advanced Computation for Architecture & Design in 2023/24 by students: Mohammad Abid, Saif Mahfouz, Ayman ElNashar and Shravya AG, and faculty: Mark Balzar and Zeynep Aksöz.