SHI YU/China Daily
Artificial intelligence (AI) models such as ChatGPT are well known worldwide. These “foundational models” have sparked worldwide excitement and interest in AI. The next step in this effort is sector-specific AI models that can be tailored to specific industries.
Such models can also be used to address development challenges such as climate change. For example, the AI-powered solution 'Agrepreneur' is an AI-powered Agrepreneur platform that provides real-time farm management advice to smallholder farmers, including how to optimize available resources and prevent crop diseases. has been developed. It also uses machine learning algorithms to assess trustworthiness.
Machine learning is also used to help farmers predict the amount of produce needed for a crop, allowing them to streamline the procurement process.
“Viamo” is also a solution that utilizes AI. Available through voice calls, farmers in areas with limited or no internet connectivity can receive guidance on sustainable farming practices. Using natural language understanding technology and pre-trained large-scale language models, text-to-speech and text-to-speech capabilities, farmers can get critical information from the app.
ClimateGPT is another example: it was trained on multidisciplinary research to provide users with a holistic understanding of climate change.
Sector-specific AI models are particularly useful in Asia and the Pacific, which are highly susceptible to extreme weather events due to climate change. Users can also respond on a small scale by creating models that tell them the likelihood of flooding or drought occurring in a specific area at a specific time.
But how can this technology be used to improve the resilience of our planet?
The advantage of sector-specific models is that their requirements are relatively minimal, unlike other tools that require millions of dollars in initial capital, human resources, and government support.
It is possible to build sector-specific models, whether you develop as an individual or as a representative of an institution. Although your requirements can be high depending on the size of your dataset and the complexity of your model or application, you can build a simple model by training the model using a laptop and a small dataset.
An AI model can be built through a series of steps, including defining the scope and purpose, establishing clear goals and parameters, and preparing and dividing the training data into smaller units. You can then customize it using platforms and frameworks like GitHub and Hugging Face, rather than starting from scratch.
Following customization, you can train the model on your data, continuously evaluating and fine-tuning it using feedback mechanisms and metrics to ensure accuracy and consistency. Multiple iterations may be required to optimize the model response. A beta testing phase involving different user groups can be used to validate the model's functionality and check for bias. This improves the reliability of the model before widespread deployment.
Depending on the complexity of the model, available resources and data, and programming language proficiency, it can take anywhere from hours to weeks to be able to deploy a working climate model.
AI, like any technology, has drawbacks and risks. We need to adhere to responsible AI frameworks that are currently being developed and mainstreamed. Over the next year or two, smaller models will begin to function under these responsible AI frameworks, similar to what we've seen with guardrails around e-commerce, social networking, and the internet's dark web. Masu.
Users developing their own sector-specific models should be aware that AI models are highly data-dependent. Poor quality data means poor quality analysis. Additionally, data bias can be reflected in what AI models produce in bulk. For example, using gender-neutral data to train AI models can be detrimental to women, who face unique challenges during a crisis.
AI serves as a power multiplier for development agencies to do their work. In the context of climate change, sector-specific AI models are being used to accelerate progress on climate action, adapt to a changing climate and solve contemporary adaptation problems, and improve the overall impact of climate change across the planet. can be reflected.
Depending on the model, this technology can be used to show global trends, or technology developed for one country can be customized for use in other countries. The presence or absence of these models can be the difference between resilience and vulnerability to climate change.
As we harness the potential of AI, specialized models for addressing common challenges like climate change offer promising paths. Once developed, these models could become a key aspect of AI solutions that set a new course for planetary sustainability and resilience.
The author is Director, Digital Innovation and Architecture, Information Technology Division, Asian Development Bank.
The views expressed are those of the authors and do not necessarily reflect the views of ADB, its management, its board of directors, its members, or China Daily.
If you have specific expertise or would like to share your opinion about our articles, please send your articles to opinion@chinadaily.com.cn and comment@chinadaily.com.cn.
