AI & Environment: How much value is embedded in carbon?

Decoding AI's Carbon Quotient: Money, Bioacoustics, and the Quest for Efficiency

Welcome (back) to AI Minds, a newsletter about the brainy and sometimes zany world of AI, brought to you by the Deepgram editorial team.

In this edition:

  • ⚙️ AI emits carbon, but how much value do we get in return?

  • 🐳 How Deep Learning helps the environment via bioacoustics

  • 🏎️ Is improving AI efficiency going to help the environment?

  • 🌍 TED Talk from Hugging Face’s Sasha Luccioni

  • 🐦 Twitter’s thoughts on AI and the environment

  • 💸 How AI, money, and carbon emissions relate

  • 🤖 Bigger isn’t always better for LLMs… 

Thanks for letting us crash your inbox; let’s party. 🎉

We coded with the brand-new Whisper-v3 over the past week, and the results were not what we expected. Check it out here!

🐎 AI & the environment: Does tech do more harm than good?

🧑‍🔬 The arXiv take…

ML developers know that AI models emit carbon. To curb the environmental impacts, many have proposed a simple solution: Make the models more efficient. After all, the less time these models spend computing, the less power they consume and the fewer emissions they make.

The authors of the paper below argue, however, that efficiency isn’t enough… What do you think?

🎥 Video: HuggingFace’s Sasha Luccioni takes on carbon emissions

“Replacing a smaller, more efficient model with a larger language model emits 14 times more carbon for the same task, like telling [a] knock-knock joke.”

Sasha Luccioni, PhD. (HuggingFace)

🐝 Social media buzz

Some say AI only adds to the world’s Carbon emissions. Others argue that we don’t have enough information yet to say for certain. Above, you’ve seen researchers’ formal calculations as well as others’ more qualitative analyses.

Here’s what Twitter (read: X) says, as bluntly and candidly as usual:

🤖 Additional bits and bytes

Interested in this topic? Check out these additional bits and bytes to delve deeper into environmental AI!

  • Environmental cost of AI - This article features similar content to the final section of AI Minds #6, (see here). In it, we analyze how environmental consequences go hand-in-hand with financial impacts. It costs money to emit carbon, so perhaps limiting carbon emissions is, in a way, economically wise.

  • Bigger isn’t always better  - On the note that Luccioni left us with with respect to small models versus big models, Andy Wang reveals that bigger isn’t always better even when it comes to performance.