Podcast Summary: "We’re doing AI all wrong. Here’s how to get it right"

Speaker: Sasha Luccioni
Show: TED Talks Daily
Airdate: October 30, 2025
Location: TED Countdown event, New York (in partnership with the Bezos Earth Fund)

Overview

In this episode, sustainability scientist Sasha Luccioni delivers a compelling talk challenging the dominant narrative around artificial intelligence (AI). Arguing that the current approach to AI development is shortsighted and unsustainable, Luccioni calls for a paradigm shift—away from ever-larger, energy-hungry models controlled by a handful of corporations, towards smaller, more sustainable, decentralized, and human-focused AI systems. With vivid examples and powerful comparisons to big oil’s legacy, Luccioni explores how we must rethink AI for the benefit of people and the planet.

Key Discussion Points & Insights

1. The AI Hype and What’s Overlooked

• The polarized conversation about AI—either as a savior or existential threat—misses the real issue: the unsustainable way AI is being developed.
• Quote [03:44]:

  “In my opinion, both of these statements are wrong. And what they do is they distract us from the real issue at hand. We’re doing AI wrong at the expense of people and the planet.”
  — Sasha Luccioni

2. Big AI’s Unsustainable Playbook

• Major tech companies (Meta, OpenAI, XAI) are pumping vast resources into gargantuan data centers (e.g., Meta’s planned data center the size of Manhattan, OpenAI’s Stargate data center).
• The environmental toll is staggering:
  • OpenAI’s Stargate expected to emit as much CO₂ as the country of Iceland.
  • XAI faces lawsuits over health and environmental impacts in South Memphis.
• Quote [05:01]:

  “Remember Big Oil? Well, now we have Big AI following the exact same playbook, using more and more resources, building bigger and bigger data structures, and selling us the narrative that this is somehow inevitable.”
  — Sasha Luccioni

3. Bigger Isn’t Always Better: The Problem with Large Language Models (LLMs)

• Current “bigger is better” AI mentality encourages more compute, more data, more energy—the de facto mantra of Silicon Valley.
• LLMs (like ChatGPT) are general-purpose and require massive energy, often for trivial tasks.
• Study shows that asking simple questions (e.g., "What's the capital of Canada?") to an LLM uses up to 30 times more energy than a smaller, task-specific model.
• Quote [07:23]:

  “Today we use AI as if we were turning on all the lights of a stadium just to find a pair of keys.”
  — Sasha Luccioni

4. The Rise and Promise of Small LLMs

• A new generation of “small LLMs” demonstrates similar performance at a fraction of the ecological and computational cost.
• Example: Hugging Face's small LM models, trained on carefully curated, mostly educational content, run locally—reducing misinformation, privacy risks, and dependence on massive data centers.
• Quote [09:13]:

  “These models are flipping the script on the bigger is better mentality... above and beyond environmental impacts, they also have benefits when it comes to cybersecurity, data privacy, and sovereignty.”
  — Sasha Luccioni

5. AI Beyond LLMs: Real-World Sustainable Solutions

• Specialized models—requiring far less energy—tackle climate change directly:
  • NASA-funded Galileo models: support crop mapping, flood detection without specialized hardware.
  • Rainforest Connection: small AI models on old solar-powered cell phones detect illegal logging via bioacoustics.
  • Open Climate Fix: uses AI to predict renewable energy output from solar and wind installations via satellite data.
• Quote [10:54]:

  “There are so many other approaches in AI that use less energy and still are really useful in our fight against climate change.”
  — Sasha Luccioni

6. Transparency and Tools for Sustainable Choices

• Lack of transparency: users do not know the energy/carbon impact of AI tools, making informed sustainable decisions impossible.
• Luccioni’s “AI Energy Score” initiative tests over 100 open source AI models, rating them on energy efficiency (1–5 stars).
• Example: To answer “What's the capital of Canada?”, Smalllm uses 0.007 watt hours, while Deep Seq uses 150 times more.
• Quote [12:20]:

  “But sadly, big AI companies didn’t want to play ball and evaluate their models... because the truth might only make them look bad.”
  — Sasha Luccioni

7. Policy, Accountability, and Taking Back Control

• There are virtually no binding policies making AI companies accountable for their energy/resource use.
• The EU AI Act introduces voluntary disclosures, but global regulation will take time—“time we simply don’t have, given the speed and the scale of the climate crisis.”
• Luccioni urges listeners: Don’t assume today’s trajectory is inevitable; we can still choose a better AI future.

Memorable Quotes & Moments

• [03:44] “We’re doing AI wrong at the expense of people and the planet.”
• [05:01] “Now we have Big AI following the exact same playbook [as Big Oil].”
• [07:23] “We use AI as if we were turning on all the lights of a stadium just to find a pair of keys.”
• [09:13] “Small LLMs... run literally on your phone or in your web browser, giving you access to state-of-the-art AI... without needing massive data centers.”
• [12:20] “Big AI companies didn’t want to play ball and evaluate their models... the truth might only make them look bad.”
• [13:21] “With every prompt, every click, and every query, we can reinvent the future of AI to be more sustainable together.”

Important Timestamps

• [03:44] — Opening argument: The real problem with AI.
• [04:23] — Big data centers & environmental/corporate examples (Meta, OpenAI, XAI).
• [05:01] — Comparison to Big Oil’s legacy.
• [06:20] — Cost of “bigger is better” in AI; example of energy waste.
• [08:19] — Small LLMs as an alternative; specifics on size, training data, and advantages.
• [10:25] — Broader applications: crop mapping, illegal logging detection, renewable energy forecasting.
• [11:52] — AI Energy Score project for model efficiency.
• [12:30] — Lack of policy and urgency for legislation.
• [13:21] — Closing call-to-action: collective agency in shaping AI’s future.

Conclusion & Takeaways

Sasha Luccioni’s talk is a thought-provoking critique and rallying cry: We need to fundamentally rethink how, for whom, and at what cost we build AI. The dominance of ever-larger, resource-intensive LLMs is not inevitable or sustainable. By prioritizing transparency, policy intervention, open-access technology, and energy-efficient models, we can—and must—shape an AI future that serves humanity and protects the planet.

Final call-to-action [13:21]:

“With every prompt, every click, and every query, we can reinvent the future of AI to be more sustainable together.”
— Sasha Luccioni