New technologies, such as rapidly advancing deep learning models, have given rise to increasingly sophisticated artificial intelligence models. With promises ranging from autonomous vehicles to highly specialized information retrieval and creation, the possibilities seem limitless.
However, there are potential dangers, such as the displacement of some work positions, privacy concerns, and concerns about materials and energy. On the latter, two specialists shared their reflections.
Impact of reliance on AI on computing infrastructure
Deep Jariwala, assistant professor of electrical and systems engineering at the University of Pennsylvania School of Engineering and Applied Sciences and Benjamin C. Lee, professor of electrical and systems engineering and computer and information sciences at the The same house of studies, have spoken about the impact that an increased reliance on AI computing will have as the infrastructure is developed to meet their increasing needs.
Jariwala and Lee explain that AI is a whole new paradigm in terms of function. While the first computer was built to do math that would take humans too long to calculate by hand, AI is less about crunching raw numbers and more about using complex algorithms and machine learning to train and adapt it to new information or situations. . This means that AI can extract information from larger data sets, such as the Internet, and provide results that are uncanny in many cases.
As certain products that use AI, such as ChatGPT and Bing, gain popularity, the nature of computing is increasingly based on inference. Lee notes that this is a slight departure from the machine learning models that were popular a few years ago, such as Google DeepMind’s AlphaGO, the machine trained to be the best player of the board game Go. Now, massive AI models are being integrated into daily operations, like running a search, and there are trade-offs with that.
Material and resource costs associated with AI
Jariwala warns that all the tasks our devices perform are transactions between memory and processors. Each of these transactions requires energy.
As these tasks become more elaborate and data intensive, the need for more memory storage and power begin to scale exponentially. Regarding memory, an estimate from the Semiconductor Research Corporation, a consortium of all major semiconductor companies, posits that if we continue to scale data at this rate, which is stored in memory made of silicon, we will exceed the global amount of silicon produced every year. This means that very soon we could reach an overdemand for this resource.
In addition, energy production is also a worrying issue. The infrastructure required to support the growing demand for AI requires a lot of energy, which can have a significant impact on the environment. The growing need for data processing and memory storage can also have an impact on the carbon footprint of the technology.
Possible solutions for diagnosed problems
A possible solution for the described context, according to Jariwala, is to investigate memory and data processing alternatives. Instead of relying on memory made of silicon, we can explore other materials that may be more power efficient, such as ferroelectric-based memory, which is a form of memory that can retain information even when power is interrupted. He also mentions the possibility of investigating processors based on more energy efficient materials.
In addition, the specialist outlines the possibility of exploring energy efficiency in data center infrastructure. Data centers are big consumers of energy, so energy efficiency in these places can make a big difference. Some strategies to improve energy efficiency include the implementation of more efficient cooling techniques, the implementation of heat recovery technologies, and the adoption of renewable energy.
Lee, for his part, indicates that it is important to equally consider individual and collective responsibility to reduce our impact on the environment. Instead of relying on the latest devices and technologies, we can choose to repair and upgrade our current devices instead of buying new ones. We can also take steps to reduce our energy use, such as turning off devices when we’re not using them and using energy more efficiently.
With the increasing reliance on AI and AI-related technologies, significant implications in terms of energy and materials are being revealed. However, as Lee and Jariwala commented, there are potential solutions to promote more sustainable practices in technology.
