Ethical Impacts of Generative AI
In the past, new technologies have shifted the balance of power between economic and cultural stakeholders.
This can hurt people socially and economically – and these negative effects can be exaggerated when technology is adopted quickly.
Generative AI technologies are changing fast now and we can already identify harm it has caused. We can also see ways that its benefits are distributed inequitably.
Below is an incomplete list of references and examples.
- What is missing from this list?
- How is it likely to change in the future?
Energy and Water
- I Live 400 Yards From Mark Zuckerberg's Massive Data Center
- Optimizing Cooling Efficiency in Modern Data Centers
- Cooling systems in data centers account for roughly 30% to 40% of total energy consumption.
- Global data center energy consumption is expected to double by 2030
- AI is Draining Water from areas that need it most
- The Environmental Impact of AI: A Case Study of Water Consumption by Chat GPT
- GPT-3 reportedly consumed approximately 700,000 liters of water during its training phase.
- Five Ways to Improve Data Center Energy Efficiency and Cut Energy Costs
- 1% of all global energy is consumed by data centers (2022)
- Data centers are expected to use 1/5th of the world’s power supply by 2025.
- Holistically Evaluating the Environmental Impact of Creating Language Model (arxiv)
- How Hungry is AI? Benchmarking Energy, Water, and Carbon Footprint of LLM Inference (arxiv)
- A single short GPT-4o query consumes 0.42 Wh,
- 700 million GPT-4o queries/day results in:
- Electricity use comparable to 35,000 U.S. homes
- Freshwater evaporation matching the annual drinking needs of 1.2 million people
- Carbon emissions requiring a Chicago-sized forest to offset
TESCREAL
- On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? (2021)
- In 2020, Google fired researcher Timnit Gebru, when she tried to publish this paper that was critical of large language models. The paper discusses environmental costs, bais, uniterpreteability and cultural homogeneity in the training data all shown to be prevelant in LLMs.
- The TESCREAL bundle: Eugenics and the promise of utopia through artificial general intelligence (pdf)
- Timnit Gebru - Eugenics and the Promise of Utopia through AGI (YouTube)
- Critical research on contemporary "AI Philosophy" movements: transhumanism, Extropianism, singularitarianism, (modern) cosmism, Rationalism, Effective Altruism, and longtermism
- Identifies connections between founders of contemporary AI movements and the founders of engenics
- Identifies overlap in the goals of contemporary AI movements and eugenecists
- Effective Altruism Promises to Do Good Better. These Women Say It Has a Toxic Culture Of Sexual Harassment and Abuse
- Karen Hao - Empire of AI
- 2025 book about OpenAI and the "AI Arms Race"
- Michael Geoffrey Asia - The Emotional Labor Behind AI Intimacy
AI Slop
- Daniel Stenberg - AI slop attacks on the curl project (YouTube)
- The I in LLM stands for intelligence Curl has a "Bug Bounty" which is overwhelmed by AI Slop
- A security but report that *does not invoke any *curl* code
- The viral AI avatar app Lensa undressed me—without my consent
- "My avatars were cartoonishly pornified, while my male colleagues got to be astronauts, explorers, and inventors"
- This is how AI Image generators see the world (Washington Post)
- AI Image Generators inherit bias from the training process
- Concept Artists Say Generative AI References Only Make Their Jobs Harder
Scams
- Microsoft Digital Defense Report 2025
- "AI-automated phishing emails achieved 54% click-through rates compared to 12% for standard attempts" p.37
- Extortion and ransomware drive over half of cyberattacks p.52-
- Generative AI is useful for cyber attack augmentation and automation.
- "Threat actors can now automate vulnerability discovery, malware generation, and data analysis. In response, defenders are also leveraging AI to detect vulnerabilities, automate patching, and improve threat intelligence."
- Woman Conned Out of $15K After AI Cloned Her Daughter’s Voice in Terrifying Scam
- "Deepfakes" aid identity impersonation crime
Finally
- AI isn't useless. But is it worth it?
- "But I find one common thread among the things AI tools are particularly suited to doing: do we even want to be doing these things?"
- Pauline Oliveros - Quantum Improvisation: The Cybernetic Presence (1999)
- Oliveros responds to the Ray Kurzweil's assertion in The Age of Spiritual Machines: When Computers Exceed Human Intelligence, that "In a hundred years there may be no clear distinction between humans and computers"
- "What would I want on a musician chip if I were to receive the benefit of neural implant technology? What kind of a 21st Century musician could I be?"