2024. 9. 19.
Go to the previous episode
In our previous discussion, we explored how artificial intelligence technology, contrary to expectations, is exacerbating the climate crisis. Now, let's examine the efforts being made to address this issue.
What's Being Done?
Fortunately, many research institutions and companies recognize the severity of this problem and are exploring various solutions. The main approaches can be categorized into three types:
Utilizing Eco-friendly Energy Sources: Efforts are underway to replace the massive amount of electricity required in the AI field with eco-friendly energy sources. By utilizing renewable energy such as solar and wind power, we can reduce dependence on fossil fuels and consequently mitigate climate change.
Efficient Operation of Data Centers: Technologies are being introduced to more efficiently use the vast amounts of freshwater that previously evaporated, through high-efficiency cooling systems and energy-saving processors. For example, Meta has implemented a system that reduces water usage by cooling air with cold water and then using this air to cool the data center interior, effectively reusing water multiple times. Google, in its Netherlands data center, is using treated wastewater as coolant, supplying 10 billion liters of cooling water annually.
Improving AI Power Efficiency: Researchers are developing new algorithms and architectures that consume less power than existing AI models. This allows for maintaining equivalent performance while reducing data center usage, thereby decreasing environmental impact and cutting costs simultaneously.
On-Device AI: A New Paradigm
The recently prominent 'on-device AI' technology presents new possibilities in addressing the climate crisis. This technology performs AI computations on individual devices instead of in data centers, offering the following advantages:
Cost reduction through independence from data centers
Diverse AI utilization possibilities without relying on networks
Providing personalized AI experiences with enhanced data security
The independence from data centers, in particular, is becoming an attractive option for tech companies in terms of solving energy consumption issues and reducing costs. This is having a significant impact, especially on mobile devices and edge devices operating with limited resources.
Conclusion
Various solutions are being proposed and rapidly implemented to address the impact of AI on the climate crisis. On-device AI technology, in particular, is advancing quickly thanks to the efforts of several tech companies. For example, some innovative companies are developing technologies to miniaturize high-performance AI models for efficient operation on mobile devices. Among these, companies like ZETIC.ai are showcasing solutions that not only miniaturize models but also maximize the performance of NPU chips within devices, offering powerful AI capabilities while improving energy efficiency.
However, what's most crucial is ongoing public interest and transparent information sharing. Considering that many companies are not transparently disclosing the usage of data centers for AI models and the amount of energy consumed, active interest and information demands from users are important.
No technology can be perfect, and AI is no exception. What we should be most wary of is not the negative aspects of technology itself, but the ignorance and indifference to its effects. To enjoy the benefits of AI technology while creating a sustainable future, we must continue our efforts to find a balance between technological advancement and environmental protection.
