Key Takeaways:
- Innovative Model: Cropin Technology’s ‘akṣara’ is the sector’s first purpose-built open-source Micro Language Model for agriculture, licensed under Apache 2.0, with no restrictions.
- Strategic Focus: Initially, ‘akṣara’ will focus on nine major crops across five countries in the Indian subcontinent, targeting key staples in the Global South.
- Resource Efficiency: Utilizing advanced compression techniques, the model minimizes environmental impact and computational demands.
- Accessibility and Empowerment: This open-source model is designed to empower agronomists, researchers, and farmers, particularly in underrepresented regions, by providing essential agricultural insights.
- Ethical AI Use: Cropin is committed to ethical AI practices, aiming to reduce biases and promote sustainable agricultural methods.
Cropin Technology, a global leader in Agtech innovations, has officially launched ‘akṣara’, the first open-source Micro Language Model (µ-LM) dedicated to climate-smart agriculture. Developed on the foundation model Mistral, ‘akṣara’ aims to revolutionize how agricultural stakeholders in the Global South access and utilize digital technologies for farming.
Features and Capabilities
Purpose and Design
“Akṣara is designed to address the unique challenges faced by the underserved farming communities by removing knowledge barriers and empowering stakeholders to develop scalable AI solutions,” stated Krishna Kumar, Founder & CEO of Cropin. This model provides contextual, factual, and actionable information, which is crucial for tackling issues like food security, climate change, and resource conservation.
Coverage and Application
The model initially covers paddy, wheat, maize, sorghum, barley, cotton, sugarcane, soybean, and millets, which are significant for the food security of the population in the Global South. “These crops are staples for billions of people, and optimizing their production is critical in our fight against food insecurity,” Kumar added.
Technological Innovations
Developed and hosted on Hugging Face, ‘akṣara’ utilizes QLoRA (Quantization and Low-Rank Adapters) for model compression, reducing its size from 16-bit to 4-bit and significantly lowering the requirements for compute power and storage. This model outperformed the GPT-4 Turbo by 40% in relevance on a randomly selected test dataset using the ROUGE scoring algorithm.
Implementation and Impact
Data Integrity and Refinement
The model was fine-tuned with over 5,000 high-quality question-response pairs specific to agriculture, ensuring the responses are both relevant and precise. “By grounding ‘akṣara’ in high-quality data and expert knowledge, we ensure that the model’s outputs are not only accurate but also practically applicable,” explained a Cropin data scientist.
Addressing Climate Vulnerability
Farmers in the Global South are particularly vulnerable to the impacts of climate change, such as erratic rainfall and extreme weather conditions. ‘akṣara’ aims to mitigate these challenges by providing insights that can adapt farming practices to changing climatic conditions, thereby enhancing resilience and sustainability.
Ethical Considerations and Future Directions
Commitment to Responsible AI
Cropin has implemented guidelines from Google’s People + AI Guidebook and collaborated with Google’s Responsible AI team to ensure that ‘akṣara’ adheres to ethical AI practices. “Our goal with ‘akṣara’ is to promote the use of AI in sustainable agriculture, focusing on practices that conserve soil, water, and biodiversity, and ensuring equitable benefits for all stakeholders,” said Kumar.
Community Engagement and Expansion
The open-source nature of ‘akṣara’ facilitates its accessibility and utility to a broad audience, including agronomists, scientists, and on-ground agricultural workers. Cropin plans to continuously update the model with more crops, languages, and regional data to enhance its applicability and effectiveness.
Cropin Technology Latest News
Cropin’s Journey to B Corp Certification (2024/02/29)
Cropin Technology has become a Certified B Corporation™, reflecting its commitment to high social and environmental standards. Its technology, deployed in 103 countries, is helping to reimagine agriculture through digitalization. Cropin’s B Corp certification underscores its dedication to sustainability and its leadership in the global ‘Ag-intelligence’ movement.
Banks a USD 13.7M Funding With Google’s Participation (2023/01/09)
Cropin Technologies, an agritech company with headquarters in Bengaluru, has secured $13.7 million from a variety of new and existing investors. The business will use the new funding largely to develop its newly released intelligent agricultural cloud platform. The business has now raised a total of $46 million in equity capital since the most recent injection. Data-driven farming is made possible by Cropin’s farm management software and apps, which link the many agri-ecosystem players. They offer users real-time data and insights to enhance agriculture’s financial and operational elements. The company asserts that it has collaborated with more than 250 business-to-business (B2B) clients and that through digitizing 16 million acres of agriculture, over 7 million farmers have benefited.
Image provided by Cropin
