Coursera Spring Sale
40% Off Coursera Plus Annual!
Grab it
Learn to design and implement a Carbon Impact Agent that monitors and reduces the environmental footprint of AI applications while maintaining business value in this conference talk from NDC London. Explore the significant environmental costs of AI development, where computational power requirements have doubled every 3.4 months since 2012, and discover how training models like GPT-3 generates over 550 tons of COâ‚‚ equivalent to 300 round-trip flights between New York and San Francisco. Understand why downstream AI application footprints remain largely overlooked despite their substantial impact, and examine how most AI designs prioritize SOPs and data quality while treating sustainability as an afterthought. Master the architecture of a multi-agent system that tracks, analyzes, and recommends strategies to actively reduce environmental impact through compute usage monitoring, energy mix analysis, task-level telemetry, and carbon intensity tracking by region. Discover practical implementation techniques including agent orchestration for monitoring and analytics, caching repeated requests, nudging users toward greener prompts, scheduling tasks during off-peak hours, and routing workloads to cleaner-energy regions. Gain insights into measuring per-task carbon costs, workflow-level emissions, and team-level benchmarks, then learn to act on these insights through UI nudges, greener deployment choices, prompt optimization, and intelligent workload scheduling to build AI systems that embed environmental accountability into intelligent applications by design.
