Sustainable development and AI

To fight climate change we need improved action.

What are the fundamental benefits using Open AI, related applications, Open Source LLMs you can recommend and why ?

Let the discussion begin !

1 Like

I’ll start,

OpenAI uses Azure infrastructure to inference, train and develop, which has been carbon neutral since 2012.


This the beginning. To understand that the data centre is net-zero. But then , if we think Scope 3, it is all community that needs to be net-zero including our travel to work for example.

Value chain needs to be net-zero for emissions. Is this reachable and how we measure and report this ?

About Scope 1,2, 3 emissions:


It’s a great question, I don’t know if such documentation exists, if not I am certain it is in the pipeline, if I can get any further details I’ll post them here.


Measuring and calculating carbon emissions are a can of worms, I don’t know how to describe it better without going into specifics, but I’ll just say that I generally don’t trust the numbers these calculations produce.

1 Like

Consider the high resource usage of OpenAI API & other LLMs. Azure might be carbon-neutral, but I doubt their entire supply chain is (though didn’t check), and probably neither your own stack is. Plus, next to carbon emissions, water consumption is another issue. Also consider client-side resource usage. After all, ChatGPT & Co. are probably unlikely to be environmental irrelevant at all. If MS does not produce 100% of its electricity on its own, it might buy up carbon-neutral electricity markets in some regions and force other consumers to use carbon-negative alternatives (not checked).

Some practical considerations for reducing energy usage through OpenAI that come to my mind include:

  • Avoid proactive content generation but favor a “generate now” button wherever possible
  • Cache responses
  • Use prompting, max_tokens or stop words to keep generated contents short
  • Benchmark and monitor your token usage and implement rate limits per user/task
  • Use cheaper models (which I assume to be also more energy-efficient)
  • For queries that are not urgent, schedule them at time windows or in regions where energy demand is lower or availability is greater

Just some brainstorming from a nonprofessional, add your ideas or corrections below.

1 Like

Details here, water positive by 2030 Net-Zero waste also by 2030 Carbon Neutral by offsetting and reduction since 2012.


Well, that’s still a couple of years from now. Plus, “water positive” measured worldwide does not provide much help to specific dry regions when more water is replenished at the other end of the world instead.

Companies are today followed by satellites so emissions reports that have false data may lead into sactions against the company involved. Here is on such service

The content question is; does Open API deliver content to ALL questions it receives so that is accountable to preventing climate change ?

is it ”weighted” toward Earth sustainability?

How we know this? How about when we move to AGI?

Plan ; going to holliday in Rome? Can I fly now, can I fly 2035 when H2 Airbuses offer net-zero footprint. Should I take a car?

Sorry for any provocative impression here,

I have no answer myself

1 Like

I’m aware, just a quick reminder that satellite based emission detection only works for companies that have a chimney, a datacenter will have zero signature to detect on such systems. :laughing:


AI carbon emissions are indeed lower than those of humans when it comes to tasks like writing and illustrating. A study by Cornell University found that an AI writing a page of text emits 130 to 1500 times less CO2e than a human doing the same


Similarly, an AI creating an image emits 310 to 2900 times less CO2e per image compared to a human illustrator

.These findings are significant in the context of environmental sustainability, as AI systems can potentially help reduce greenhouse gas emissions. However, it’s important to note that emissions analyses do not account for social impacts such as professional displacement, legality, and rebound effects

1 Like

Interesting argument, but with all due respect I think their calculation of human text writing emissions is pretty much bs. :slight_smile:

To calculate the carbon footprint of a person writing, we consider the per capita emissions of individuals in different countries. […] Assuming that a person’s emissions while writing are consistent with their overall annual impact, we estimate that the carbon footprint for a US resident producing a page of text (250 words) is approximately 1,400 grams CO2e.

It’s not like people did not breathe, eat, or commute if they were not writing text … Rebound effect definitely kicks in here, we are just going to producing more and more text and throw away lots of it.

I could see Open AI supporting sales and marketing teams for the customers of ClimateTech SaaS providers. End users - homeowners - purchase solar, heat pumps and energy efficiency solutions. But they only do so once persuaded to - typically by a sales rep who presents the economic, independence and sustainability benefits.

Could Open AI significantly decrease the sales cycle through more persuasive messaging? Are there new ways for consumers to take advantage of climate tech that we don’t yet see?