AI's dirty little - clean little - secret

 

There has been much talk lately about ChatGPT and other generative AI systems, most of it centered on its ability to let students generate essays, to take journalism jobs away, to help or hurt almost every industry. 

But little has been said about the net effect of these systems in terms of their impact on climate, via the increased computing power that’s needed to power such systems. 

My attention to this was drawn by an article – theoretically written by a human – on Wired.com.  The article is called “The Generative AI Search Race Has a Dirty Secret”.  That page also provides an audio snippet of about 6 minutes, well worth a listen.

As I was about to write the post about how AI is about to boost the carbon footprint of computing, I did a minor amount of further research and found this article which talks of the promise of AI in terms of reducing carbon footprint and in general being somewhat of a savior in terms of climate change.

The article opens with an interesting quote:

 ”unbound by the restrictions of common sense, readily accessible models today build on semantic relationships derived from large bodies of text, or relationships between large corpuses of text and images.  The next generation of models already being built will include video and 3D modeling.  Soon, we may hope to design houses by starting with phrases like “single-story mid-century modern, south facing on a forested hillside, three bedrooms”. 

The article goes on to consider what happens when we (should) add elements of recycled materials, or design for lower energy use into the description of the desired building.  There’s other research that indicates a positive effect for AI in terms of carbon reduction, such as this one from the World Economic Forum.

That article has the same quandary as I did when I started the “dirty secret” blog post, saying:

·  By 2030, the tech could help cut global greenhouse gas emissions by 4%, according to a recent study by accounting firm PricewaterhouseCoopers for Microsoft.

·  But there are energy consumption concerns around AI, algorithms and the processing of large amounts of data.

So… who could I ask about this?  What sort of wise system could provide an answer about the net effect of generative AI systems?

Oh, I know: a generative AI system!

So I asked Chat GPT this question:

Overall, since generative AI will cause an increase in carbon footprint, yet has the potential to lower carbon footprints due to innovative design and other AI applications that reduce the carbon footprint, what do you predict will be the net effect of generative AI with respect to carbon footprint?

Here is the answer given by ChatGPT:

It's difficult to make a definitive prediction on the net effect of generative AI on carbon footprint, as it will depend on a variety of factors, including how the technology is developed and used, and the extent to which it is adopted.

On the one hand, the energy consumption required to train large-scale generative AI models can be significant, and this could contribute to an increase in carbon footprint. Additionally, if the applications of generative AI are focused on industries that are particularly carbon-intensive, such as transportation or manufacturing, then the net effect could be an increase in carbon footprint.

On the other hand, there are a number of potential applications of generative AI that could have a positive impact on carbon footprint, such as improving the efficiency of energy systems, optimizing the use of renewable energy, and helping to reduce waste and emissions. If generative AI is used in these ways, it could help to reduce carbon footprint and mitigate the effects of climate change.

Overall, it's likely that the net effect of generative AI on carbon footprint will depend on the balance between these different factors, and how the technology is developed and adopted in the coming years. It will be important for developers and policymakers to prioritize environmentally responsible applications of generative AI, and to ensure that the energy consumption associated with the technology is sustainable and powered by renewable sources.

When I asked the System specifically about the “dirty” and “clean” aspects, I got more detail:

Will generative AI increase the carbon footprint of computing?

Generative AI has the potential to increase the carbon footprint of computing, depending on how it is implemented and used.

Training a generative AI model requires significant amounts of computing power, and this can contribute to increased energy consumption and carbon emissions. In particular, large-scale models like GPT-3 require massive amounts of computational resources, including high-end hardware like graphics processing units (GPUs) and tensor processing units (TPUs), which can consume a lot of energy.

Additionally, the use of generative AI in applications like chatbots, virtual assistants, and automated content generation can also contribute to increased energy consumption if these applications are widely used and require a lot of processing power.

However, there are also efforts underway to make generative AI more energy-efficient, such as developing more efficient hardware and software algorithms. Furthermore, there is a growing interest in using renewable energy sources, such as wind and solar power, to power data centers where AI models are trained and hosted.

Overall, while generative AI does have the potential to increase the carbon footprint of computing, there are also measures that can be taken to mitigate its impact on the environment.

Will generative AI increase the carbon footprint of computing?

Generative AI has the potential to increase the carbon footprint of computing, depending on how it is implemented and used.

Training a generative AI model requires significant amounts of computing power, and this can contribute to increased energy consumption and carbon emissions. In particular, large-scale models like GPT-3 require massive amounts of computational resources, including high-end hardware like graphics processing units (GPUs) and tensor processing units (TPUs), which can consume a lot of energy.

Additionally, the use of generative AI in applications like chatbots, virtual assistants, and automated content generation can also contribute to increased energy consumption if these applications are widely used and require a lot of processing power.

However, there are also efforts underway to make generative AI more energy-efficient, such as developing more efficient hardware and software algorithms. Furthermore, there is a growing interest in using renewable energy sources, such as wind and solar power, to power data centers where AI models are trained and hosted.

Overall, while generative AI does have the potential to increase the carbon footprint of computing, there are also measures that can be taken to mitigate its impact on the environment.

Do generative AI systems have the potential to help solve the climate crisis?

Yes, generative AI systems have the potential to help solve the climate crisis in a variety of ways. Here are some examples:

1. Climate Modeling: Climate models are used to make predictions about the future of our planet, such as changes in temperature and precipitation patterns. Generative AI can help improve the accuracy of climate models by processing large amounts of data and identifying patterns that might be difficult for humans to detect. This can help researchers make more accurate climate predictions and identify potential solutions to mitigate the effects of climate change.
2. Energy Optimization: Generative AI can be used to optimize energy usage in buildings and industrial processes, reducing energy waste and greenhouse gas emissions. By analyzing data from sensors and other sources, generative AI systems can identify areas where energy use can be optimized, such as by adjusting temperature and lighting settings or identifying equipment that is consuming too much energy.
3. Renewable Energy: Generative AI can help optimize the deployment and management of renewable energy resources, such as solar and wind power. For example, AI systems can predict energy production from solar panels based on weather patterns, helping to better integrate solar power into the grid and reduce reliance on fossil fuels.
4. Carbon Capture: Carbon capture technology involves capturing carbon dioxide emissions from industrial processes and storing them underground. Generative AI can help optimize the design of carbon capture systems, making them more efficient and cost-effective.

Overall, generative AI systems have the potential to help solve the climate crisis by providing more accurate climate modeling, optimizing energy usage, facilitating the deployment and management of renewable energy resources, and improving the efficiency of carbon capture technology. However, it's important to ensure that the development and use of AI systems are done in a way that is environmentally responsible and sustainable.

What do you (humans) think?  What role do project managers play?  What role do project LEADERS play?  Personally, I think it’s a big, important role, and it all starts (as did this post) with asking the right questions.  Please comment, if possible, without using ChatGPT, but if you do, say how you used it!