Uncover the hidden environmental costs of large language models and learn actionable strategies to balance AI innovation with a more sustainable future.
Like the monolith from 2021: A Space Odyssey, AI is being touted as something that will herald a new era of computational power and human capability. The things that AI is already able to do are truly mind-blowing - and it’s still very much in its infancy.
But let’s take the rose-tinted glasses off for a moment. We are all staring down the barrel of a full-blown, man-made climate crisis. Technology is helping us mitigate its impacts somewhat, with improvements in fields like renewable energy offering glimmers of hope.
AI holds the potential to make our lives easier and more efficient. However, AI’s relationship with energy consumption is complicated. At worst AI is a huge energy and resource hog - at best, AI professionals are working on mitigating that fact.
Pinning down the exact carbon impacts of AI, and especially of specific Large Language Model (LLM) AIs like ChatGPT, is tough. But energy use - and misuse - is totally under human control. It’s our responsibility as humans who use AI tools to do so ethically and mindfully.
Before we explore the ways we as AI practitioners can redress this much-needed balance, we need to fully examine some of the ecological worries inherent when using AI, especially Large Language Models.
4 Major Environmental Concerns with Large Language Model AI Usage
1) LLM’s Massive Energy & Resource Consumption
AI - and especially elements like deep learning and neural networks - requires much more computational power than standard computing. The sheer amount of information storage and data crunching inherently needed by AI consumes far more energy than any kind of computing that has ever come before.
The wide-reaching, multi-purpose AI tools that make the headlines, like ChatGPT and Gemini, are called “large language models”. To vastly simplify, these tools maintain access to trillions of data points that they can draw upon when they give a response, and contain countless parameters to help them score and link those data points; hence why their responses seem so natural and human.
But maintaining access to those countless data points, parameters, and the inferred links between them - and growing its access to millions more data points every day - takes a lot of energy, infrastructure, and hardware. And that’s before we get to the energy consumption involved in answering the countless requests they receive every 24 hours.
This becomes especially concerning when you zoom out and think about the vast data centres needed to operate cloud tools of all kinds, not just AI ones. Data centres constantly consume energy. Their servers are running 24/7, as are the routers and switches that get data where it needs to go. IT equipment needs to be cooled, so air conditioners are always running. Their employees have to commute to and from them every day, likely using combustion engines of some kind.
With energy supply around the world still so reliant on fossil fuels, this is understandably a significant draw on carbon resources. (Though it’s worth noting that both Google and Microsoft have made strides in their efforts to become carbon neutral/negative.)
Research from University of Massachusetts, Amherst found that merely training a large language AI model emits approximately the same amount of CO2 as 5 average cars across their whole lifespan. That’s just training the thing - once the LLM is actually put to use, it could end up consuming orders of magnitude more energy than that throughout its existence.
2) Mining Rare Earth Minerals for Tech Manufacturing
In order to build new computer hardware, manufacturers need to mine rare earth elements (REEs) and minerals. This strips the Earth of its natural finite resources; and once they are mined, they still need to be transported, processed, and then manufactured into whatever board or device they are destined to become - all of which may still rely on fossil fuels and polluting practices.
Also well worth mentioning here is the continued labour and human rights abuses taking place throughout technology supply chains.
Heading back to AI for a moment, computer hardware undergoes quite intense usage when running large language models day in, day out. This all chips away at the hardware’s natural lifespan. When that hardware fails and needs to be replaced, that perpetuates the demand for more resource stripping and potentially sends more e-waste to landfill.
3) AI’s Trendiness and Simplicity Leads to LLM Overuse
Sadly, many view AI as the latest tech toy. And who can blame them? Interacting with publicly available LLM AI tools like Chat GPT is deceptively simple. With similar effort to a Google search, users can generate new text, stories, art, or code; get answers to questions; research and hash out ideas; and much more.
As such, both individuals and businesses alike are already using LLM AI tools alarmingly frivolously, such as asking their LLM of choice what the weather is up to or what’s the best air fryer to buy - things that would usually be the subject of a normal web search.
And though typing a prompt into a generative AI tool may appear very similar to a web search, each LLM AI query comes at a far higher carbon cost than anything generated by standard computing. Research suggests that a ChatGPT query consumes around 60 times more energy per query than a simple Google search.
But this isn’t just a case of ease - it’s a case of trendiness too. Organisations want to appear ahead of the curve by adopting AI and LLM tools, because it’s the hot new tech. So some end up doing so just for the sake of it; without a meaningful use case that justifies the need for such a complex, heavy, and potentially polluting computational tool.
4) How You Code - And Where You Compute - Matters
In computer programming, there's usually more than one way to code a solution to a problem. Sometimes code is kept efficient and lean, only carrying the bare minimum instructions needed to carry out the functions in question. Other times, code becomes bloated and inefficient, full of exceptions and weird sticking plaster workarounds.
When code is poorly optimised, it takes more energy and computing power to navigate the inefficient twists, turns, and dead ends. Code that is well-optimised is far more energy efficient.
This is true of any software, not just tools that rely on AI. But if your software does involve AI, you need to make sure your AI is targeted precisely to your specific use case and the code doesn’t make erroneous or frivolous AI requests.
Simply moving data around also consumes energy. A data packet travelling from LA to London, through various routers and data centres, is going to have a larger carbon impact than one going from Liverpool to Manchester. So think: does a computation have to happen on some distant server somewhere and then be transported to you? Or can that process take place within your own network? Or perhaps directly on the relevant device? There’s no need in making your data move further than it has to!
Ways AI is Already Helping Fight Climate Change
AI itself isn’t some world-ending carbon hog - it’s how we use it that counts.
In fact, it’s already being used to great effect in the fight against climate change. Let’s acknowledge some of the good work that is already happening in the eco-AI space:
- Eco-friendly Number Crunching - AI tools are already being used in waste processing, in reforestation efforts, and even measuring changes to icebergs. AI is also being used to optimise energy use and distribution, monitor ocean health, and water conservation - with scope for uses like precision agriculture.
- Accurate Weather & Climate Event Predictions - AI weather forecasting tools can predict weather with more accuracy than standard weather simulation systems. Experts have argued for the use of AI to advance climate modelling and prediction.
- Waste Intelligence - British company Grey Parrot supports companies with AI-powered waste analytics, enabling facilities to “recycle more and waste less”.
- Smart Urban Planning - AI shows a lot of promise in areas like city planning, infrastructure design, predictive climate modelling, and generally helping to create more sustainable, enjoyable cities.
5 Considerations for Eco-Friendly AI Adoption
There are many sensible ways SMEs can benefit from AI and be as kind as possible to the planet. Here are a few things you can bear in mind before embarking on your next AI tech transformation project:
1) Consider Use Case First, AI Novelty Last
We get it, AI is trendy. Companies want to boast about how their tools are using the latest technology buzzword as it makes them seem cutting edge. Yet this trendiness can lead to AI solutions being applied in situations where standard computation would have worked equally well.
Our advice? Maintain a single-pointed focus on the specific use cases you need from your new technical solution and what problems you need it to solve. If that involves AI, then great! We’re here to help. But don’t try to shoehorn in a particular kind of tech where it might not be needed.
2) Model Size Should Fit the Scope of the Project
Not all AIs are created equal. Yes, the large language models make the headlines, but not all AI implementations require such massive, wide-reaching datasets.
Small language models (SLMs) are artificial intelligence tools which rely on much smaller, more targeted data sets. This makes them far narrower in scope, but also often far more carbon efficient than their LLM cousins.
SLMs can be perfect for applications that require some of the more flexible and artificially creative elements of AI, but only need a limited focus; for example an AI-powered website chatbot that only needs to know details about a company’s product catalogue.
In contrast, using a vast LLM like ChatGPT to power a simple website chatbot would be like using a sledgehammer to crack a nut!
3) Be Aware of How AI Differs from Standard Computing
Large language, generative artificial intelligence is very different from any kind of computing that has come before. It requires a huge amount of data crunching, which in turn requires a lot of hardware and energy resources to maintain.
Keep this in mind when deciding to use AI, and don’t get carried away with using overpowered LLM AI solutions when a SLM or more conventionally programmed solution would suffice.
But here’s one thing you and your teams can do in the here and now: don’t overuse generative LLM tools for frivolous things where a simple web search would suffice!
4) Aim for Efficiency in All Digital Transformation
Whether your solution uses AI or not, aim to design the most energy and computationally efficient solution. For our more tech savvy readers, this might mean optimising your code so it doesn’t run into errors, exceptions, and dead ends.
But creating an efficient solution truly starts at the ideation phase. Does that data point really need to be fetched from the other side of the planet? Do all of our networked devices need to be set up to carry out this complex type of calculation, or will one or two machines suffice? Does the software really need to speak to an AI in order to do [X]? Do we really need new hardware when the old stuff is still up to spec?
5) Keep Your Tech Supply Chain Clean
We wholeheartedly welcome the fact that an increasing number of data centre operators are committing to use only renewable energy in their data centres. Alas, despite these noble efforts, it still doesn’t mean that the whole supply chain is rendered magically spotless.
The tech hardware supply chain is reportedly slow to address human rights abuses. Rare earth elements are expensive and polluting to extract - and only available in finite quantities.
Sadly this means that no tech supply chain is ever going to be 100% clean. But here’s another tip you can start right now: research your supply chains and do what you can to ensure suppliers align with your principles.
In Conclusion
Right now, AI definitely helps to expand human throughput, which is a great thing. But, despite the hand wringing in the headlines and the shareholder-pleasing statements coming out from "Big AI", machines cannot think like humans yet, and can’t make decisions on what is right or wrong. They’re not even close.
AI certainly magnifies our ability to make positive impacts in the world. But it magnifies our ability to make negative impacts too. Whether those negative impacts are cruelly negative on purpose, or come about from well-meaning misunderstanding or error.
This is why it's so important for us as decision-making humans to carefully consider the impacts of our use of technology before we rush into using AI simply because it's the latest tech trend.
Thankfully, AI is still in its infancy, and we have time to make it a much less polluting, efficiency-driving force for good.
But it’s up to us as humans and decision makers to create that future.
If you foresee a use for responsible, sustainable AI in your future technical transformation projects, book a totally free consultation call with our expert team today.
Related reading
As we reach the end of the calendar year, our glasses fill with our favourite Christmas tipple, and we slow down for the Christmas break, it's a great opportunity to reflect on the past year. Way back in January, I sat down to make my top 5 predictions for AI in 2024, and whilst i've kept my eye on how each was progressing... now it's time to dig in and assess how those predictions really fared.
1️. Slowing of the Generative AI Hype Train
Prediction: It was one of the noisiest advancements of 2024, butorganisations have struggled to consistently reap the promised rewards.Generative AI is not going anywhere but expect the hype to cool to a more considered approach as those that have watched from the sidelines dip their toes in and proceed with caution until issues with security, regulatory uncertainty, and legal precedent are cleared up.
Verdict: Unclear. (But largely incorrect.)
Generative AI (or GenAI) remains one of the most controversial technologies of recent years, and measuring the hype train canlead to different answers depending on where you look or who you ask.
In terms of adoption, a CEPR survey highlights that the adoption of GenAI has accelerated much quicker than technologies such as the PC or the Internet. Yet, on the other hand, surveys such as this one from Boston Consulting Group highlight the alarming statistic that 74%of companies struggle to achieve and scale value with GenAI.
The views are similarly conflicting when you turn to investment, with plenty of talk of a GenAI “bubble” emerging, whilst simultaneously TechCrunch highlighting $3.9 billion of investments in GenAI startups secured in Q3 of2024, alongside OpenAI’s whopping $6.6 billion round, valuing the company at $157 billion.
Whether the hype train is slowing or not is likely to dependmore on your own optimism about GenAI as a technology. The one thing we can sayfor certain: GenAI isn’t going away, particularly whilst investment remains strong. Those that identify the right strategy and use cases for it will find themselves in a strong and enviable position ahead of those who are struggling to get value from their investments.
2️. LLMs vs SLMs
Prediction: As developers begin to understand the strengths of Large Language Models for conversational interfaces, we're likely to see far more chat bot style solutions for interacting with users. The demand for competitiveness in this space is likely to drive competition with Small Language Models, as leaner and less intensive language models close the performance gap on their larger counterpart.
Verdict: Correct.
The year started strongly for the Small Language Models (SLM), as numerous big players in the Large Language Model (LLM) market continued to emphasise and expand their smaller model offerings (Llama, Phi, Mistral), and IBM returning to the LLM market with its Granite models focusing on smaller size to lower cost without losing performance. But the battle didn’t stop at SLMs,with developers competing to release so-called Tiny Language Models (<1 billion parameters) such as HuggingFace’s SmolLM series, and even Super Tiny Language Models, with the aim of significantly reducing the compute costs required to run language solutions and opening the opportunity for language interfaces to be embedded within low-spec hardware.
Beyond just parameter size, there has been substantial critique of the model “scaling laws” often cited by LLM developers as areasoning for continually growing model size to increase performance, with evidence beginning to suggest this isn’t as clear-cut as previously claimed, and a number of studies (such as DeepSeekLLM) challenging this narrative further. (Interestingly, OpenAI itself has seemingly acknowledged this shift with the release of its o1 model, focusing on increased inference compute rather than a growth in parameter size)
3️. The Rise of Organisational Knowledge Bases
Prediction: To leverage the potential of recent AI advances, organisations are likely to be forced to look inwards at their own data infrastructure. Expect a further push in Digital Transformation to create internal Knowledge Bases to power AI applications, as well as a rise in the use of Knowledge Graphs to ground these in fact.
Verdict: Correct – mostly?
The understanding that AI needs high quality data to be truly successful, alongside the realisation that allowing third-party providers to train models off your data risks leaking of sensitive information has led many organisations to review their data infrastructure, with those with already accessible data able to demonstrate quick wins on AI workflows to increase buy in. Projections for the cloud market continue to show strong growth, where organisations such as Microsoft have invested heavily in cloud infrastructure across the globe (Mexico, Japan, Malaysia, and many more), a commitment matched by competitors such as Amazon (UK, Italy, Japan, and more).
Whilst there has been a drive from data infrastructure to provide AI ready features, there is little evidence to suggest Knowledge Graphs have been considered important in this, however there are early signs of recognition in the importance of graph structures in GenAI workflows. GraphRAG emerged as a technique to improve the retrieval and output capabilities of genAI solutions, with Microsoft continuing to research the benefits of this approach. Meanwhile, Knowledge Graph market leader neo4j saw its revenue grow beyond $200 million this year, crediting its ability to improve accuracy, transparency and explainability of GenAI solutions a major factor to this growth.
Whilst enterprise scale graphs may not be disrupting the market yet, expect them to become a hot topic as GenAI adoption grows.
4️. Large Vision Models
Prediction: As the Large Language Model momentum stables out, expect to see competition to develop foundation models in the Computer Vision space. These models will be able to form a wide range of tasks on image and video data, helping accelerate adoption of AI for visual tasks.
Verdict: Correct – but quietly.
The Computer Vision market has been steadily growing, and it Is projected to continue, but Large Vision Models (LVMs) haven’t received anywhere near the level of attention and hype of LLMs despite the range of use cases they enable. Likely this is because these models are mostly being picked up by designers and developers, rather than being exposed directly to consumers. However, early in the year, we saw image capabilities rolled into popular consumer facing tools.
After a late 2023 announcement, OpenAI rolled out its GPT-4 vision model throughout 2024, a key advancement to their offering, allowing ChatGPT to run off so-called Multi-Modal Models (models that can process a range of input types such as text, image, audio and video). In the last couple of years,Multi-Modal models have become the new frontier for the major AI model developers as they seek to combine previously separate processing streams into one model architecture. Across the board numerous new models have been coming quick and fast: Meta, Anthropic, Google, Amazon, Mistral, and more have all made major Multi-Modal releases this year.
Another major advancement this year has also come in the form of Generative Video, with Runway and OpenAIs Sora catching the headlines at different stages of the year.
Aside from model-development, product developers have sought to integrate computer vision progress into their solutions, and buyers have been following suit with continued growth of adoption in Computer Vision tools for security, health and safety, and quality control amongst other use cases. In fact a Gartner survey this year predicted 50% of warehouse operations will leverage a Computer Vision solution by 2027, emphasizing that competition in the Large Vision Model space is likely to grow.
5️. Cloud Computing vs Edge Computing
The growth in data and AI applications is putting a tremendous strain on cloud compute infrastructure, and the growth of distributed technologies looks to bea further headache. Expect to see more Internet of Things technologies focused on computing on device and Federated Learning, whilst cloud providers seek to find ways to reduce cost and latency to remain competitive.
Verdict: Unclear.
The main computing takeaway from 2024 is that both Cloud and Edge computing are growing quickly and as demand for hardware and compute power surges there has been little need for the two markets to directly compete. Whilst we touched on cloud growth earlier, Edge Computing growth is arguably stronger, with increased demand in Real-time analytics, automation, and enhanced customer experiences.
The Edge trend is unlikely to slow down, and chip manufacturers are investing heavily in preparation, with NVIDIA, AMD, Intel and Altera (amongst others) all making significant improvements to their Edge AI offerings as they continue to posture over a market that is likely to grow substantially in the coming years. Whilst Cloud and Edge are co-existing peacefully at present, this remains a space to watch in the future as awareness of the advantages of Edge computing could muscle in on Cloud solutions.
In Conclusion
The AI market in 2024 has been a filled with noise, confusion and chaotic energy as organisations have faced pressure to find and adopt AI use cases in a turbulent economic environment. That being said, the noise had begun to quieten in the latter portion of the year, with success stories and torch carriers beginning to emerge to guide the journey, leaving an optimistic outlook for 2025.
Ready to start your AI journey?
No matter your technological know-how, we’re here to help. Send us a message or book a free consultation call today.