Compression and complexity: Making sense of Artificial Intelligence

Compression and complexity: Making sense of Artificial Intelligence

Advertisements

Compression and complexity: Making sense of Artificial Intelligence

30 June 2023

Sergio Scandizzo

 

 

 

Sergio Scandizzo is Head of Internal Modelling at the European Investment Bank.
.

 

Artificial Intelligence (AI) is expected to have a major impact on Europe in the coming decades. Sergio Scandizzo explains how the concepts of compression, complexity and depth can help us understand the potential implications of AI for our daily lives.

ChatGPT and other instances of ‘generative AI’ have recently taken the internet by storm and, in parallel, generated a mountain of critical comments ranging from the awed and terrified to the unimpressed and disparaging. On one side is the traditional concern that AI can help people cheat, replace human judgement in key decisions for our lives and ultimately damage livelihoods by raising unemployment.

On the other, somehow illogically, although perhaps as an understandable reaction to those fears, critics have tried to find fault with AI’s performance: it cannot solve certain mathematical puzzles (nor can the majority of humans); it writes essays that are predictable and solely based on searching the available literature (as most essays written by humans sadly are); on occasion, it can produce absurd results and reach biased and discriminatory conclusions (otherwise said, it looks as human as it gets).

So, while we tremble at the thought of a dystopic future of technological unemployment, AI-controlled governments, and stultified students, at the same time, we berate current AI applications for not yet being that kind of God-like, infallible intelligence capable of solving any possible problem without fail. The reality is that most ‘failures’ of AI – not being original, basing decisions on existing information, numbly following rules or simply failing several times at complex tasks – are typical human features.

Some critics note that ChatGPT uses the most cited texts, assuming that those are the most scientifically reliable, to come up with answers. True, but what do most people do when they write an essay? First, they read the most cited texts. Similarly, others devise tricky mathematical questions to make the programme produce wrong answers (which are the kind of answers most humans would give). I wonder, therefore, about how many university essays, honestly written by mediocre students in the future, will look like they were written by ChatGPT or some other AI engine and attract unfair accusations of plagiarism.

The objective of Deep Blue or AlphaGO is not to be intelligent, but to play Chess or Go like an intelligent being (us). That they clearly can do so is disquieting perhaps because it suggests that it doesn’t necessarily take human intelligence to play these games, even at the highest level.

The same holds true for so-called creative tasks. If a machine can write a perfectly acceptable, even if not especially original, essay, it gives us pause for thought primarily because it forces us to rethink both the value of certain products of our intelligence and the meaning we attach to them. This is presumably what makes some commentators desperate to find fault in AI’s performance, as if they were keen to reassert the primacy of human intelligence against an existential threat.

Compression as intelligence

Let us try to look at the problem from another perspective. A ‘lossy’ compression algorithm is an algorithm that saves memory space by identifying statistical regularities across a set of data and storing a single copy of patterns that recur multiple times, without being exactly the same. The results of such technique are worse than what you get using a ‘lossless’ compression algorithm, where the original information can be completely reconstructed, but good enough for several practical applications.

It works especially well with images and music, much less well, unsurprisingly, with text and numbers. For example, if a lossy algorithm will store just one copy for several similar-looking areas of a picture, the reconstructed image may become slightly blurred but would still be recognisable overall. On the other hand, if the algorithm stores only the average of several similar numbers in a spreadsheet, the results will likely be useless.

Last February, science fiction author Ted Chiang wrote a very thoughtful piece in which he argued that ChatGPT works very much like a lossy algorithm applied to the internet, whereby it samples a large amount of information and repackages it in the form of text that is not exactly the same as any of the texts available online, but close enough to look both correct and original.

Aside from the fact that he may have stumbled across a definition applicable to a lot of what passes for creativity these days, what is especially intriguing is the use of compression as a metaphor for intelligence and specifically, his observation that the best way to devise a way to efficiently compress a set of data is to understand them.

Indeed, if we need to compress, for instance, the Fibonacci sequence, which is an infinite series in which each number is the sum of the previous two, we would do well by storing only three equations – F0 = 0 (applies only to the first integer), F1 = 1 (applies only to the second integer), and Fn = Fn-1 + Fn-2 (applies to all other integers) – rather than a very long sequence of integers hoping that the next user will guess the rule.

Complexity and depth

In a different context, Nobel laureate Giorgio Parisi argues that the problem of finding the simplest description of a complicated set of data corresponds to finding the scientific laws of the world and is “often taken as a sign of intelligence”. To clarify this idea, Parisi draws on the concept of the algorithmic complexity of a string of symbols.

The latter is defined as the length of the shortest computer programme producing that string as an output. In the Fibonacci sequence example, such a programme will incorporate, in the simplest possible fashion, the three equations above, thereby obtaining a very short description of an (infinite) sequence. On the other hand, if we examine the string “Dkd78wrteilrkvj0-a984ne;tgsro9r2]3”., nm od490jwmeljm io;v9sdo0e,.scvj0povm]]-” the shortest programme most likely will have to look like:

Print (once):

‘Dkd78wrteilrkvj0-a984ne;tgsro9r2]3”., nm od490jwmeljm io;v9sdo0e,.scvj0povm]]-‘

This is longer than the string to be printed. Equally important, however, is the concept of the logical depth of an algorithm, which is the actual amount of CPU time needed to execute it. In the Fibonacci case, while the algorithm is short, the actual amount of CPU required to execute it is potentially infinite as the sequence goes on forever. In the random string above, the algorithm is indeed not very efficient with respect to the length of the string, but its execution is very quick.

We say therefore that the former algorithm has low complexity and high logical depth, while the latter exhibits the opposite features. A good scientific theory has low complexity (it gives the simplest explanation of data) but potentially high logical depth (it explains a lot and may require a very long time to compute all its implications).

As an example, E=mc2 has very low complexity, but an enormous level of logical depth as it implies a very profound and far-reaching set of results. It follows, alas, that in many practical cases we settle for more approximated theories with higher complexity and lower logical depth either because of efficiency (approximated theory may be good enough for certain tasks) or because we simply cannot afford the CPU time required (Parisi gives the example of when we meet a lion on our path and we must make a decision in real time or else become its dinner).

This trade-off between complexity and depth is fundamental in understanding how intelligence, human or otherwise, works, yet most discussions of AI seem to ignore it. ChatGpt may well be, as Chiang says, an imperfectly compressed version of the available data, but so is most of our learning. Intelligence, amongst other things, is the ability to perform those somewhat imperfect compressions that balance cognitive objectives with our natural constraints.


NB: This article gives the views of the author, not the position of EUROPP – European Politics and Policy, the London School of Economics or the European Investment Bank. Featured image credit: Emiliano Vittoriosi on Unsplash


Read LSE‘s

.

 

 

How AI physics has the potential to revolutionise product design

Advertisements

How AI physics has the potential to revolutionise product design

This article was published by the World Economic Forum on 26 June 2023.
Joris PoortFounder and Chief Executive Officer, Rescale

  • Artificial Intelligence (AI) tools like ChatGPT have become mainstream, but the use of AI in the product design process is less widely known.
  • Engineering and scientific computing are harnessing AI to define a new era of innovation across industries.
  • Computer simulations are increasingly using AI to understand how different product designs will perform.

 

The impact of Artificial Intelligence (AI) on digital services is quickly becoming apparent. Tools like ChatGPTBard, and GitHub Copilot are transforming how we work and live.

What is less known is the potential for AI to revolutionise the basic nature of research, engineering, and physical product design.

Engineering and scientific computing is now the foundation of innovation. These methods often require massive computing power from supercomputing clusters (also known as high-performance computing or HPC) to run detailed simulation models that replicate the real world.

Across industries, research and development (R&D) teams use digital simulations to explore the physical world effectively. Use cases vary from inventing life-saving medicine, improving aircraft design and pioneering sustainable energy to creating self-driving vehicles, and refining manufacturing processes, among many other possibilities.

Now, AI offers the potential to supercharge engineering and scientific computing and transform how organizations innovate.

Making R&D more efficient

The computer simulations used today for engineering and scientific computing are increasingly benefiting with additional assistance from AI (and in some cases replaced by AI), dramatically lowering costs and helping engineers find the best answers faster.

Running simulations can be expensive, often requiring supercomputers to crunch massive data sets and execute highly complex calculations. But if you can build a machine learning (ML) model on how the physics works, you don’t need to run simulations every time since your ML inference model can extrapolate the answer from the data.

This means you get the answers on how a particular design will perform faster and cheaper. Over the long term, machine learning, physics-informed neural networks, and other AI-based tools will become standard tools for all engineers and scientists to maximize productivity within R&D development.

The era of AI-assisted engineering

Organizations that become adept at creating well-crafted AI-physics models will gain a solid competitive advantage. Such capabilities will help them establish a fundamental physics understanding of how different product designs will perform in real life. This has huge implications for how organizations retain knowledge in their research and product development.

AI-physics models can capture best practices knowledge about how physical objects behave – information that traditionally has been in the head of an expert, such as a scientist, engineer, or designer.

An aircraft engineer, for example, has accumulated knowledge about the best design approaches for the shape of wings, which then informs their choices regarding the types of design options they explore with digital simulations.

But that kind of information can now be captured with AI, which can then come up with a shortlist of design suggestions that engineers can then further explore with digital simulations.

Critically, suppose a company can create an ML model of best practices for wing designs (by training the AI tool on its body of knowledge about airplane wings). In that case, it can retain that expertise, regardless of if an engineer leaves the company.

This also brings far greater agility to an organization. If a company wants to build a new kind of plane that is more stable in high winds, an ML application can quickly generate the best options for the shape of the wings, helping the organization rapidly spin up new prototypes to enter new markets.

AI will physically shape our world

Given AI’s ability to help us understand the physical world, we are likely to see new shapes in all types of products, from buildings and aircraft to furniture and automobiles. Such innovations are being driven by another variant of AI: generative design.

Generative design works much like generative AI for writing text. By providing some basic guidance (prompts) about what you are trying to design, generative design tools will output many possible options, some of which you would not have thought up on your own.

By letting the software decide the design based on your performance objectives, some fascinating possibilities can result. Generative design, for example, is creating prototypes with a very biological look.

Generative AI for product design evolution and optimization example of a load-bearing bracket. Source: Rescale.

Navigating the AI transition for science and engineering

Organizations that embrace AI will accelerate engineering and scientific discovery while developing innovative new solutions that would be computationally prohibitive using traditional approaches.

Despite the promise of AI, organizations across industries will need to establish engineering and research best practices to help ensure they navigate this transition safely to maximize the benefits to society without needless risk.

Most importantly, AI is only as good as the information it trains on. Organizations will still need to do much work to provide the essential information to make the AI tool smart in the right ways.

Also, legal issues for AI are still very much undefined. Organizations must carefully review the outputs of AI to ensure accuracy, as well as watching for any ethical red flags.

Security is also another important consideration to make sure AI practices don’t accidentally expose intellectual property or proprietary information.

Certainly, guardrails for how organizations use AI are essential as we work through the early days of this new technology. But with some thoughtful measures in place, AI can safely open up all new possibilities for research and development, helping organizations move faster, become more agile, and discover better ways to invent the future.

Supporting the adoption of AI physics will help us make better products faster, accelerate the R&D innovation process, and explore the boundaries of knowledge to develop new engineering breakthroughs and scientific discoveries.

Exploring the Intersection of AI and Sustainable Architecture

Advertisements

The image above is for illustration and credit to GettyImages.

AI and the Built Environment: The Next Generation of Design Solutions

Exploring the Intersection of AI and Sustainable Architecture

Artificial intelligence (AI) has been making waves in various industries, and it’s no surprise that it’s now finding its way into the world of architecture and design. As we strive to create more sustainable and efficient buildings, AI has the potential to revolutionize the way we approach the built environment. By exploring the intersection of AI and sustainable architecture, we can unlock the next generation of design solutions that will help shape the future of our cities and communities.

One of the most significant ways AI can contribute to sustainable architecture is through the optimization of building design. Traditionally, architects and engineers have relied on their experience and intuition to create energy-efficient buildings. However, AI algorithms can analyze vast amounts of data and consider numerous design variables to identify the most sustainable and cost-effective solutions. This data-driven approach can lead to more innovative designs that minimize energy consumption, reduce waste, and lower the overall environmental impact of buildings.

For example, AI can be used to optimize the orientation, shape, and size of a building to maximize natural light and minimize heat gain. This can result in a more comfortable indoor environment while reducing the need for artificial lighting and air conditioning. Similarly, AI can help architects select the most appropriate materials and construction techniques to improve a building’s thermal performance and reduce its carbon footprint.

Another area where AI can make a significant impact is in the design of urban environments. As cities continue to grow and urban populations increase, there is a pressing need to create more sustainable and livable urban spaces. AI can be used to analyze complex urban systems and identify the most effective strategies for improving air quality, reducing traffic congestion, and promoting walkability and public transportation. By using AI to inform urban planning decisions, we can create more sustainable and resilient cities that are better equipped to face the challenges of the future.

In addition to optimizing design, AI can also play a crucial role in the ongoing management and maintenance of buildings. By integrating AI with building management systems, it’s possible to monitor and analyze the performance of a building in real-time. This can help identify inefficiencies and potential issues before they become significant problems, allowing for more proactive maintenance and reducing the overall environmental impact of a building throughout its lifecycle.

Furthermore, AI can be used to create more responsive and adaptive buildings that can adjust to changing conditions and occupant needs. For instance, AI-powered systems can learn from occupants’ behavior and preferences to optimize lighting, heating, and cooling, resulting in a more comfortable and energy-efficient environment. This level of personalization can not only improve the overall user experience but also contribute to greater sustainability by reducing energy waste.

As we continue to explore the intersection of AI and sustainable architecture, it’s essential to consider the ethical implications of these emerging technologies. While AI has the potential to revolutionize the built environment, it’s crucial to ensure that these advancements are used responsibly and equitably. This includes addressing issues related to data privacy, algorithmic bias, and the potential displacement of human workers in the design and construction process.

In conclusion, AI offers a wealth of opportunities for creating more sustainable and efficient buildings and urban environments. By harnessing the power of AI, architects, engineers, and urban planners can develop innovative design solutions that minimize environmental impact, improve building performance, and enhance the overall quality of life for occupants. As we continue to explore the intersection of AI and sustainable architecture, we can look forward to a future where our built environment is smarter, more resilient, and more sustainable than ever before.

 

.

.

 

Circular Economy and Digital Technologies can

Advertisements

The Built environment needing a sustainable future should rely on a circular economy and digital technologies to support its building industry. 

These can help it to cope with its worldwide environmental challenge. How is the question that is answered by Jacqueline Cramer.

.

..

How circular economy and digital technologies can support the building sector to cope with its worldwide environmental challenge?

By Jacqueline Cramer

 

The building sector can address pressing environmental problems by leveraging two major trends: circular economy and digital technologies. Circular building practices emphasize restorative design principles, which can significantly reduce the amount of virgin material used and the environmental footprint of buildings. When combined with digital technologies, circular practices can achieve even higher environmental benefits. Such technologies enable visualization of the environmental impact along the entire value chain, facilitating smart design, production, and use to increase material- and eco-efficiency. However, realizing the full potential of these trends requires more than just technological advancements. Institutional, behavioral, and socio-economic system changes are essential to effect a transition towards a circular and digital economy. To facilitate such a transition, a new form of governance is needed, in which network governance complements conventional public governance. Network governance fosters the formation of coalitions of willing partners that jointly strive towards the goal of system change, creating a fertile ground for a new economic paradigm, behavioral change, government regulation and innovation. The effectiveness of network governance in supporting public governance depends on the specific socio-cultural and political context of a country. However, a thoughtful application of this governance model can facilitate the building sector’s journey towards greater material- and environmental efficiency.

Introduction

The building sector is confronted with the imperative of accelerating its environmental performance. Currently, building and construction generate 36 percent of global energy consumption, produce 40 percent of waste and account for roughly 40 percent of carbon dioxide emissions worldwide1. To tackle these environmental challenges, the building sector must capture the opportunity that two major trends provide: digital technologies and the circular economy. This article explains why these trends can be critical for mitigating the environmental impact of the building sector and outlines strategies for how their implementation can be achieved and accelerated.

Digital technologies

The application of digital technologies can benefit the building sector by making the building process more material- and eco-efficient2. A broad field of digital technologies are available and continuously scaling, including artificial intelligence, big data, cloud computing, cyber physical systems, blockchain and virtual and augmented reality3. However, the building sector has just begun to adopt these emerging technologies. Integrating these technologies into daily work processes would significantly add value to the sector4. For instance, data management tools—such as Building Information Modeling (BIM), material passports, lifecycle analysis and material flow analysis—can enhance transparency about the environmental performance of the entire building chain and provide insight into how the chain can become more eco-efficient5.

The broad field of virtual and augmented reality can provide a 3D understanding of how a building is constructed, with what materials, and how this can be attuned to the needs of the customer. In addition, it can optimize resource use during the construction, maintenance, and end-of-life phases. An example is the use of digital twins6. This is a virtual representation of an object or system that spans its lifecycle, is updated from real-time data, and uses simulation, machine learning and attendant reasoning to help decision-making, also about material-efficiency7. In addition, 3D printing offers a greener building technique that eliminates a great amount of CO2 emitting and energy-consuming processes compared to conventional building techniques8. Thus, digital technologies can help improve the environmental performance of buildings, particularly when combined with the circular economy.

Circular economy

The concept of the circular economy is simple yet urgent. It highlights the fact that we are overconsuming natural resources, some of which are scarce, on a global scale. In 1970, we only needed one earth to provide mankind with the necessary resources; nowadays we need 1.75 earths. If we continue on our current path, we will require 3 earths by 20509. The Circular Gap Report has revealed that our world is still largely linear10, as we only bring 8.6% of what we use back into the cycle, resulting in a Circularity Gap of over 90%. To address this issue and become more prudent with raw materials, energy, and water, pleas are made to move to a circular economy11. There have been various definitions for the term ‘circular economy’12. However, the common denominator is that it is restorative by design and aims to keep products, components, and materials at their highest utility and value, distinguishing between technical and biological cycles13. This notion is particularly significant important for the building sector because of the high percentage of waste produced. However, this sector is characterized by strong project-based institutionalized practices and market mechanisms, which in many aspects do not facilitate the inclusion of circular economy principles14.

Technically, it is possible to consume far fewer raw materials in the building sector and drastically reduce CO2 emissions. We can extend the lifespan of buildings, redesign them with circularity in mind, reuse parts of them and recycle their materials15. Three Dutch examples serve to illustrate the benefits of building with circular economy principles. For instance, the distribution system operator Alliander—an entity responsible for distributing and managing energy to final consumers—opened its new office in 2015 in Duiven. Although everything about the building exudes style and newness, almost nothing in it is actually new. In fact, 83% of the materials used in the building are recycled. Similarly, in the new Venlo town hall (established in 2016 in the Netherlands) all the raw materials used in the construction can be fully reused with no loss of value. Moreover, the town hall building is entirely energy neutral, thanks to features such as solar panels, thermal energy storage, and solar boilers. The Green House pavilion is the final example, designed to be temporary, as the municipality of Utrecht has plans to redevelop the area in 15 years. The construction used as many recycled materials as possible, which will also be reused when the building is removed. And ultimately, when that happens, there will be no trace left of The Green House in or on the land. The building’s construction is designed to ensure that no pipes, cables, or sewage will remain in the soil under the pavilion, thus minimizing its impact. However, scaling up such iconic projects and making circular building mainstream remains a significant challenge. It requires system innovation, in which technological change goes hand in hand with a socio-economic and behavioral change. The main obstacles to realizing this system change include a focus on short-term goals, complex supply chains, a lack of collaboration between stakeholders, and the absence of a commonly agreed definition of the circular economy within the industry16.

Governance

Experiences in circular economy have demonstrated that the aforementioned obstacles can be overcome with effective governance during the transition to a circular system17. This shift requires a fundamental departure from the current linear system in which products are carelessly discarded after use. No single entity, whether it be a company, local government, or NGO, can undertake such a comprehensive system change on their own. Collaboration among partners who are committed to contributing to the change is necessary to establish a robust network. To ensure its efficacy, this network should be orchestrated through a concept known as ‘network governance’. Network governance is not meant to replace conventional public governance, but rather to complement it. It facilitates the attainment of circular objectives and strengthens societal support for more stringent government measures.

A comparative study encompassing 16 countries has illustrated that network governance can offer substantial added value18. However, the extent to which network governance can support public governance is contingent upon specific socio-cultural and political contexts19. For instance, in countries where the government takes a strong leadership role in circular economy and receptivity towards network governance is high, the conditions for initiating and accelerating circular economy are propitious. The Dutch circular building examples mentioned above serve as a case in point. In contrast, where both forms of governance are weak, it is more arduous to launch circular initiatives. Nevertheless, opportunities for developing circular economy can be identified in all 16 countries studied. In Australia, for instance, industry, government, and NGOs exhibit a rather antagonistic attitude towards one another. However, this does not preclude cooperation among these actors in sectors such as building; it simply necessitates additional incentives. For example, when commissioning parties cooperate in restructuring an urban area and implementing circular strategies, they can urge the network of contractors to exchange data and adopt an integrated circular approach. Digital technologies can reinforce such cooperation.

Hence, the building sector worldwide can make substantial strides on the path to circular economy when new forms of network cooperation among pertinent actors are implemented in conjunction with government leadership. Individual actors frequently hesitate to assume leadership roles in system change, as they do not perceive it to be their core business and await others to step forward. To resolve this predicament, independent intermediaries, known as transition brokers, can play a pivotal role in orchestrating the change process. They can align actors with divergent interests around a shared vision and resolve impasses. To be effective, transition brokers must possess a specific set of competencies and acquire the mandate to function as intermediaries. Once accepted, transition brokers can accelerate the process significantly.

Researchers can also contribute to the transition towards a circular building sector. However, to render their research socially relevant, individual projects should be clustered around themes that collectively portray the broader picture of transitioning to a circular economy. In this way, research can be mobilized that centers on fundamental solutions confronting society today. Generalists with sufficient knowledge about the variety of innovations and the specifics of the building sector are certainly equipped to bundle research and highlight the most promising innovations. These knowledge brokers can facilitate the utilization of research in practical applications in the building sector, in the short or long term20. This would enhance the value of the arduous work undertaken by numerous researchers in the field of the built environment.

The image above is credit to IStock.

.Read more on NATURE urban sustainability

Is generative AI bad for the environment?

Advertisements

Is generative AI bad for the environment? A computer scientist explains the carbon footprint of ChatGPT and its cousins

By Kate Saenko, Boston University

The image above is on The Generative AI Race Has a Dirty Secret, credit to WIRED

AI chatbots and image generators run on thousands of computers housed in data centers like this Google facility in Oregon.
Tony Webster/Wikimedia, CC BY-SA

Generative AI is the hot new technology behind chatbots and image generators. But how hot is it making the planet?

As an AI researcher, I often worry about the energy costs of building artificial intelligence models. The more powerful the AI, the more energy it takes. What does the emergence of increasingly more powerful generative AI models mean for society’s future carbon footprint?

“Generative” refers to the ability of an AI algorithm to produce complex data. The alternative is “discriminative” AI, which chooses between a fixed number of options and produces just a single number. An example of a discriminative output is choosing whether to approve a loan application.

Generative AI can create much more complex outputs, such as a sentence, a paragraph, an image or even a short video. It has long been used in applications like smart speakers to generate audio responses, or in autocomplete to suggest a search query. However, it only recently gained the ability to generate humanlike language and realistic photos.

Using more power than ever

The exact energy cost of a single AI model is difficult to estimate, and includes the energy used to manufacture the computing equipment, create the model and use the model in production. In 2019, researchers found that creating a generative AI model called BERT with 110 million parameters consumed the energy of a round-trip transcontinental flight for one person. The number of parameters refers to the size of the model, with larger models generally being more skilled. Researchers estimated that creating the much larger GPT-3, which has 175 billion parameters, consumed 1,287 megawatt hours of electricity and generated 552 tons of carbon dioxide equivalent, the equivalent of 123 gasoline-powered passenger vehicles driven for one year. And that’s just for getting the model ready to launch, before any consumers start using it.

Size is not the only predictor of carbon emissions. The open-access BLOOM model, developed by the BigScience project in France, is similar in size to GPT-3 but has a much lower carbon footprint, consuming 433 MWh of electricity in generating 30 tons of CO2eq. A study by Google found that for the same size, using a more efficient model architecture and processor and a greener data center can reduce the carbon footprint by 100 to 1,000 times.

Larger models do use more energy during their deployment. There is limited data on the carbon footprint of a single generative AI query, but some industry figures estimate it to be four to five times higher than that of a search engine query. As chatbots and image generators become more popular, and as Google and Microsoft incorporate AI language models into their search engines, the number of queries they receive each day could grow exponentially.

AI chatbots, search engines and image generators are rapidly going mainstream, adding to AI’s carbon footprint.
AP Photo/Steve Helber

AI bots for search

A few years ago, not many people outside of research labs were using models like BERT or GPT. That changed on Nov. 30, 2022, when OpenAI released ChatGPT. According to the latest available data, ChatGPT had over 1.5 billion visits in March 2023. Microsoft incorporated ChatGPT into its search engine, Bing, and made it available to everyone on May 4, 2023. If chatbots become as popular as search engines, the energy costs of deploying the AIs could really add up. But AI assistants have many more uses than just search, such as writing documents, solving math problems and creating marketing campaigns.

Another problem is that AI models need to be continually updated. For example, ChatGPT was only trained on data from up to 2021, so it does not know about anything that happened since then. The carbon footprint of creating ChatGPT isn’t public information, but it is likely much higher than that of GPT-3. If it had to be recreated on a regular basis to update its knowledge, the energy costs would grow even larger.

One upside is that asking a chatbot can be a more direct way to get information than using a search engine. Instead of getting a page full of links, you get a direct answer as you would from a human, assuming issues of accuracy are mitigated. Getting to the information quicker could potentially offset the increased energy use compared to a search engine.

Ways forward

The future is hard to predict, but large generative AI models are here to stay, and people will probably increasingly turn to them for information. For example, if a student needs help solving a math problem now, they ask a tutor or a friend, or consult a textbook. In the future, they will probably ask a chatbot. The same goes for other expert knowledge such as legal advice or medical expertise.

While a single large AI model is not going to ruin the environment, if a thousand companies develop slightly different AI bots for different purposes, each used by millions of customers, the energy use could become an issue. More research is needed to make generative AI more efficient. The good news is that AI can run on renewable energy. By bringing the computation to where green energy is more abundant, or scheduling computation for times of day when renewable energy is more available, emissions can be reduced by a factor of 30 to 40, compared to using a grid dominated by fossil fuels.

Finally, societal pressure may be helpful to encourage companies and research labs to publish the carbon footprints of their AI models, as some already do. In the future, perhaps consumers could even use this information to choose a “greener” chatbot.

Kate Saenko, Associate Professor of Computer Science, Boston University

This article is republished from The Conversation under a Creative Commons license. Read the original article.