# Extended Intelligences

## Week I

#### Extended Intelligences: A Reflective and Practical Guide to AI

Artificial Intelligence (AI) is reshaping industries, education, and everyday life, but it comes with significant material, societal, and ethical challenges. This reflection explores AI's complexities by integrating course learnings, practical projects, and research insights. The goal is to offer a balanced understanding of AI's impacts, applications, and responsibilities while providing educational tools for further exploration.

<figure><img src="https://2762819000-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FhvmmscJZstMqdsz9JVoU%2Fuploads%2FoEfnA4kOjAYOtsnVJ4BG%2FAi%20scheme.png?alt=media&#x26;token=afe0c96a-e218-4b6f-856e-4bdd224b8c91" alt=""><figcaption><p>AI Structure scheme</p></figcaption></figure>

Check this clear and simple video about AI structure... 👇

{% embed url="<https://www.youtube.com/watch?v=qYNweeDHiyU>" %}

### The Environmental and Social Costs of AI

**Material Realities**\
AI systems rely on global infrastructures tied to natural resources, labor, and energy. These systems:

* Depend on **rare earth minerals** for devices, contributing to environmental degradation and electronic waste.
* Consume vast energy, with data centers accounting for about **2% of global electricity use**​.
* Rely on large-scale **water usage** for cooling data centers, impacting ecosystems.

**Carbon Footprint**\
Training large AI models, such as GPTs, generates enormous carbon emissions. For instance, training one model can emit over **626,000 pounds of CO₂**, equivalent to hundreds of flights between New York and San Francisco.

**E-Waste Crisis**\
AI systems drive hardware obsolescence, leading to rising **electronic waste**.&#x20;

Hazardous chemicals in e-waste contaminate soil and water, threatening health and biodiversity. By 2050, e-waste is projected to exceed 120 million metric tons.

<figure><img src="https://2762819000-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FhvmmscJZstMqdsz9JVoU%2Fuploads%2FY0tk26K6r37tcaqloLBC%2Fclimate%20impact%20of%20AI_Page_06.jpg?alt=media&#x26;token=02527854-248f-4cc3-84cf-194f89188b55" alt=""><figcaption></figcaption></figure>

### AI’s Societal Impacts

**A Mirror of Society**\
AI reflects the biases present in its training data. This has led to concerns about perpetuating discrimination and inequality, especially in applications like facial recognition and predictive policing​.

**Temporal Influence**\
AI risks locking society into outdated perspectives. By relying on historical data, systems may entrench biases and fail to adapt to modern challenges, such as climate change or evolving social norms.

**Opportunities for Governance**\
When integrated with **environmental, social, and governance (ESG)** principles, AI can promote transparency and sustainability. For example, AI can help monitor carbon emissions, optimize recycling systems, and advance equitable policymaking.

<figure><img src="https://2762819000-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FhvmmscJZstMqdsz9JVoU%2Fuploads%2Fu8P0BJvwxKEZjcBsKChm%2Fgenai_cartography.jpg?alt=media&#x26;token=d46aae32-0607-4ebf-9340-f4cf4d161deb" alt=""><figcaption><p><a href="https://tallerestampa.com/en/estampa/cartography-of-generative-ai/"><mark style="color:red;">Cartography of generative AI</mark></a></p></figcaption></figure>

***

## Week II

### **Collaborative Project**

**Objective**\
This project envisions a Solar-Punk-inspired AI tool designed for MDEF students in 2044. The tool assists students in identifying research paths aligned with their passions and interests by analyzing personal traits and matching them with topics in the Atlas of Weak Signals (AoWS).

**Concept Development**\
The team focused on creating a personalized, introspective AI system. Recognizing the challenges MDEF students face in narrowing their professional focus, the tool uses data-driven insights to position individuals near relevant topics in the AoWS.

### **Datasets Creation**

1. **Atlas of Weak Signals (AoWS) Dataset:**
   * Extracted data from the AoWS website, including titles, descriptions, keywords, and trends.
   * Expanded incomplete entries and formatted the dataset into a comprehensive CSV for analysis.
2. **Student Dataset:**
   * Created a survey with 54 questions covering personal aspects like goals, motivations, interests, and skills.
   * Gathered responses from 17 participants and exported them into a CSV for integration.

**Integration and Analysis**

* The datasets were combined using AI tools to map students' profiles into a latent space, clustering them with topics from the AoWS for tailored recommendations.

#### Tools and Methods

1. **all-MiniLM-L6-v2:**
   * Used for generating sentence embeddings, clustering data, and creating dimensional relationships in the AoWS dataset.
2. **pandas.DataFrame:**
   * Handled data organization, cleaning, and transformation of CSV files into usable formats.
3. **Plotly for UMAP Visualization:**
   * Created interactive 3D visualizations to explore patterns in the datasets and map students to relevant topics.

{% embed url="<https://www.youtube.com/watch?v=4-WIvmVFTc8>" %}

### Ethical and Bias Considerations

* **Bias Risks:**
  * Selection bias from using ChatGPT for dataset descriptions.
  * Cultural biases in survey questions and annotations.
* **Mitigation Steps:**
  * Included an open-ended comment box in the survey for feedback on perceived biases.
  * Conducted manual checks on AI-generated annotations for accuracy and inclusivity.

The project demonstrated the potential of localized, introspective AI tools for educational contexts. By aligning individual profiles with curated topics, the tool helps students navigate complex decision-making about their professional and creative paths while highlighting the need for ethical oversight and iterative refinement.

{% embed url="<https://docs.google.com/presentation/d/19_bl08aC6fLdF6qQ6glx7N2ASFBpge1xyM4x8RdscFg/edit#slide=id.g3150ba2326b_1_7>" %}
Presentation PathGuide AI
{% endembed %}

<figure><img src="https://2762819000-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FhvmmscJZstMqdsz9JVoU%2Fuploads%2F3uhB2DXjExaqNs4aG5pC%2Fgraph%204.gif?alt=media&#x26;token=ab21d2b2-7a6d-4448-b648-371bfabe2e7e" alt=""><figcaption></figcaption></figure>

Check the full code below 👇

{% embed url="<https://colab.research.google.com/drive/1kgnVZjqV3QWZmb1ihLZRu7IYZ4uwxCvv#scrollTo=UIOE8s5OLlIZ>" %}

## Building a Responsible AI Future

**Balancing Innovation and Responsibility**\
AI’s rapid development necessitates thoughtful consideration of its impacts. Responsible design focuses on:

* **Reducing Resource Use:** Developing energy-efficient algorithms and hardware.
* **Addressing Biases:** Actively mitigating inequalities in datasets.
* **Promoting Transparency:** Ensuring stakeholders understand the environmental and ethical costs of AI.

**Sustainability as a Goal**\
Frameworks like **solar-punk** envision a harmonious future where technology aligns with ecological stewardship. By treating AI as part of larger systems, designers can create tools that advance sustainability without exacerbating environmental harm.

***

### Glossary for AI Exploration

<details>

<summary><strong>Artificial Intelligence (AI)</strong></summary>

Tools that simulate human-like tasks using data and algorithms.\
**Data:** Information that represents human or natural activities, essential for AI operations.

</details>

<details>

<summary><strong>Data</strong></summary>

&#x20;Information that represents human or natural activities, essential for AI operations.

</details>

<details>

<summary><strong>Machine Learning (ML)</strong></summary>

A type of AI where systems learn from data to make decisions.

</details>

<details>

<summary><strong>Deep Learning</strong></summary>

Advanced ML using layered networks (neural networks) for complex analysis.

</details>

<details>

<summary><strong>Large Language Models (LLMs)</strong></summary>

AI models trained on massive datasets to generate text resembling human communication.

</details>

<details>

<summary><strong>Design Fiction</strong></summary>

The use of speculative storytelling and prototypes to explore future scenarios.

</details>

<details>

<summary><strong>Solar-Punk</strong></summary>

A movement imagining sustainable, nature-integrated futures

</details>

<details>

<summary><strong>Google Colaboratory</strong></summary>

An online tool for running Python code, popular in AI experimentation.

</details>

<details>

<summary><strong>Neural Networks</strong></summary>

AI systems inspired by the human brain to recognize patterns in data.

</details>

<details>

<summary><strong>Dataset</strong></summary>

Structured collections of data used for training AI models.

</details>

<details>

<summary><strong>API (Application Programming Interface)</strong></summary>

Tools that let software systems work together.

</details>

<details>

<summary><strong>Latent Space</strong></summary>

A multidimensional space in AI where relationships between data are abstractly represented.

</details>

***

## Reflection Conclusion 📝

Artificial Intelligence is a powerful yet imperfect tool. Its applications can transform industries and education but come with environmental and societal responsibilities. By approaching AI with critical awareness and sustainability in mind, we can leverage its potential while minimizing harm. Educational initiatives like this seminar offer invaluable opportunities to build knowledge and apply AI responsibly for a better, more equitable future.

<figure><img src="https://2762819000-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FhvmmscJZstMqdsz9JVoU%2Fuploads%2F5jbnbJmpv9ZgKEoBpEiA%2FSustainable-AI-future.jpg?alt=media&#x26;token=c720de7e-df64-460d-b4f0-a537831bdfbb" alt=""><figcaption></figcaption></figure>