616: Whats Really Going On with AI Data Centres with Dr. Victoria Plutshack

The podcast discusses the environmental and societal impacts of rapidly expanding AI infrastructure, with a focus on data centers. These facilities, essential for training and running large language models (LLMs), consume vast amounts of electricity - already accounting for half a percent of the UK's total usage - and require continuous power, making reliance on intermittent renewable sources like wind energy insufficient. Despite claims of "green" data centers, the reality is that even a moderately sized facility would demand far more energy than regional renewable infrastructure can reliably provide, potentially displacing clean energy from other areas and increasing dependence on fossil fuels. The UK and other countries are seeing government support for AI growth zones with favorable planning permissions, often driven by economic incentives and lobbying from major U.S. tech companies, raising concerns about foreign ownership and long-term environmental costs.

Further discussion highlights broader concerns about AI's disproportionate impact on vulnerable groups, particularly women. AI-generated deepfakes and non-consensual intimate images pose serious legal and psychological harms, while automated systems threaten to displace workers in administrative roles - jobs often held by older women - who may lack the resources to retrain. AI models also inherit and reinforce existing gender and racial biases from training data, potentially shaping children's perceptions through unregulated AI-generated content. The conversation underscores the importance of public resistance, with examples like proposed moratoriums on data center construction in Scotland, suggesting that local opposition could slow AI expansion and allow for more thoughtful, sustainable development. The episode also touches on the psychological toll of AI and climate anxiety, advocating for informed action as a means of empowerment.

• What if you built a local, low-energy AI tool that helps communities track proposed data center developments?

  • Move: Use open-source LLMs with RAG to build a lightweight web app that scrapes public planning registries (e.g., UK local councils) and summarizes new data center proposals, highlighting location, size (MW), and environmental claims.
  • Why Now?: Local resistance (e.g., Edinburgh's moratorium push) shows growing public concern - timely to empower grassroots actors with accessible tools before more large-scale data centers lock in. Regulatory windows are open.
  • Expected Upside: First-mover tool in civic-tech niche; could attract municipal partnerships, funding, or integration into advocacy platforms. Builds personal reputation in sustainable tech and delivers tangible community value.

• What if you transitioned your AI-powered SaaS product to prioritize energy efficiency over feature bloat?

  • Move: Audit your product's AI usage (e.g., token count, model size), then refactor workflows to use smaller models or retrieval-augmented generation (RAG) to cut compute by at least 50%, passing savings to users via pricing or speed.
  • Why Now?: Energy costs and environmental scrutiny are rising; large data centers (200+ MW) strain grids - developers who act early can differentiate on sustainability while reducing cloud bills.
  • Expected Upside: Lower operational costs, stronger branding as a climate-conscious product, and potential appeal to ESG-focused customers or investors.

• What if you launched a micro-consulting service helping small businesses avoid AI overkill?

  • Move: Offer structured audits for solopreneurs or small teams to assess whether they truly need AI (e.g., large LLMs) vs. simpler automation or curated knowledge bases; provide actionable, energy-conscious alternatives.
  • Why Now?: Market is flooded with AI-rebranded tools and panic-driven adoption - many users waste money and energy on unnecessary AI. Demand exists for sober, practical guidance.
  • Expected Upside: Establish authority in mindful tech adoption; generate leads through content marketing (e.g., case studies on energy/cost savings); create scalable templates or tools from client work.

• Audit your software projects for energy efficiency by measuring and optimizing token usage, API calls, and model size - prioritize smaller, open-source models and RAG over large LLMs for lower environmental impact.
• Host AI models locally or on green infrastructure when possible, using tools like open-source LLMs to reduce reliance on energy-intensive cloud data centers.
• Research and identify nearby data centers using public resources (e.g., APRS website) to understand local environmental impacts and inform infrastructure decisions in your area.
• Advocate to local policymakers by contacting council members or representatives to support moratoriums or stricter regulations on unchecked data center expansion, especially in energy-sensitive regions.
• Monitor and limit high-energy AI features in your products - avoid implementing energy-intensive functions like bulk image generation or massive code rewrites without cost and environmental impact assessments.