a solarpunk ai manifesto
technology for earth, community, and future
AI Today: A Turning Point
Current Path: monolithic AI models, massive carbon emissions, resource depletion.
Result: acceleration of ecological collapse.
Truth: this does not have to be inevitable. We can choose a different future (Strubell, Ganesh, & McCallum, 2019). It will depend on what we choose to use this technology for.
Social Media, Advertising, and AI: A Hidden Crisis and a Bad Choice
Today, social media is one of the largest consumers of AI and AI technologies..
- Over 80–90% of the user experience is now AI-mediated: feeds, recommendations, ads, moderation, notifications, etc. (Nature Communications, 2022).
- Billions of AI inferences occur every second, consuming massive continuous energy globally.
- Social media companies optimise for attention, not well-being, ecology, or community.
Key Environmental Problems:
- Energy drain: Billions of micro-predictions running 24/7, powered by nonrenewable energy grids.
- Water usage: Data centre cooling for social platforms uses millions of gallons per year.
- E-waste: Accelerated device life cycles are partly driven by social media demand.
- Carbon emissions: Cumulative operational carbon load across Meta, TikTok, Google, and other centralised social media platforms are huge.
Key Social Problems:
- Psychological exploitation: AI predicts and manipulates emotional states to maximise platform engagement (Nature Communications, 2022).
- Opportunity cost: AI talent and compute are siphoned into optimising ad clicks, not healing the planet.
- Centralised control: Power and profit are concentrated among a few monopolies at enormous ecological and social cost.
Social media represents one of the worst ecological and ethical misuses of AI resources in its short history.
Green AI Principles: Another Way
- Local First: Deploy lightweight, decentralised models in the periphery.
- Renewable Roots: Train and run AI only on 100% renewable energy (Microsoft, 2024).
- Small is Beautiful: Prefer efficient, specialised models over bloated monoliths.
- Open Access and Commons: Share AI knowledge freely for ecological good (Climate Change AI, 2019).
- Carbon Consciousness: Publicly disclose AI carbon, energy, and water footprints.
From Dirty AI to Green AI
| Dirty AI | Green AI |
|---|---|
| Centralised, opaque, carbon-heavy | Decentralised, transparent, renewable |
| Designed for capitalist incentives | Designed for community and ecology |
| Endless growth at all costs | Regenerative balance with the biosphere |
| Controlled by few | Controlled by many |
Our Future: A Solarpunk AI World
- AI managing solar microgrids for villages and communities (Wildlabs.net, ongoing).
- AI to assist in regenerating forests and monitoring oceans (Wildlabs.net, ongoing).
- Citizen scientists protecting biodiversity using AI (Wildlabs.net, ongoing).
- Community-owned AI training sustainably with a minimal footprint.
- AI-assisted housework and light farm or garden work.
- AI-assisted coding and technological development frameworks.
- AI-assisted labour (mental/physical) to free up time with the specific goal of spending more time outdoors and having more leisure time.
- Decentralised social networks and digital gardens unoptimised by AI.
Reallocate the Intelligence
Today, vast AI resources are wasted on:
- Addictive content feeds
- Manipulative ad targeting
- Attention mining for profit
At the cost of:
- Ecological collapse
- Psychological exploitation (Nature Communications, 2022)
- Social fragmentation
Primary Data Sources for Chart:
- Stanford Institute for Human-Centered Artificial Intelligence (2025)
- McKinsey & Company (2024)
- Vention AI Adoption Statistics (2024)
- Exploding Topics AI Trends (2025)
- National University AI Statistics Report (2025)
Solarpunk Reallocation Vision
| Wasted On | Reallocated Toward |
|---|---|
| Infinite Content Scrolls | Local renewable energy grids |
| Ad Auctions and Surveillance | Ecosystem regeneration and biodiversity monitoring |
| Outrage Algorithms | Climate crisis prediction and adaptation systems |
| Vanity Filters and Data Mining | Global knowledge commons for regenerative design |
| Attention Traps | Community empowerment, education, and collective health |
Closing Remarks
We must appropriate machines not be appropriated by them.
— Guattari (1989)
The very notion of the domination of nature by man stems from the very real domination of human by human.
— Bookchin (1982)
Technology and AI must become a tool for ecological, social, and mental regeneration, not capitalist exploitation. Decentralised AI can help to democratise its use and mitigate environmental catastrophe. Generalist and monolithic AI will need to be broken up into an ecosystem of specialised and decentralised machines. In principle, this network of machines will still be able to perform and answer general inquiries while also helping us to limit energy consumption, water usage, and planned obsolescence as environmental factors.
References
Bookchin, M. (1982). The ecology of freedom: The emergence and dissolution of hierarchy. Palo Alto, CA: Cheshire Books.
Climate Change AI. (2019). Tackling climate change with machine learning. Retrieved from https://www.climatechange.ai/
Exploding Topics. (2025). 54 NEW Artificial Intelligence Statistics (Mar 2025). Retrieved from https://explodingtopics.com/blog/ai-statistics​
Guattari, F. (1989). The three ecologies. London, UK: The Athlone Press.
McKinsey & Company. (2024). The state of AI in early 2024: Gen AI adoption spikes and starts to generate value. Retrieved from https://www.mckinsey.com/~/media/mckinsey/business%20functions/quantumblack/our%20insights/the%20state%20of%20ai/2024/the-state-of-ai-in-early-2024-final.pdf​
National University. (2025). 131 AI Statistics and Trends for 2025. Retrieved from https://www.nu.edu/blog/ai-statistics-trends/
Nature Communications. (2022). Social media, machine learning and human behavior. Retrieved from https://www.nature.com/articles/s41562-022-01321-3
Stanford Institute for Human-Centered Artificial Intelligence. (2025). The 2025 AI Index Report. Retrieved from https://hai.stanford.edu/ai-index/2025-ai-index-report
Strubell, E., Ganesh, A., & McCallum, A. (2019). Energy and policy considerations for deep learning in NLP. Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, 3645–3650. https://doi.org/10.18653/v1/P19-1355
Vention. (2024). AI Adoption Statistics 2024: All Figures & Facts to Know. Retrieved from https://ventionteams.com/solutions/ai/adoption-statistics
Wildlabs.net. (Ongoing). WILDLABS AI conservation technologies. Retrieved from https://wildlabs.net/