As the world progresses into the future, AI, specifically Generative AI (GenAI), is increasingly becoming an integral part of our daily lives. However, with this rapid advancement, there is a growing need for stringent global regulations to govern its use. In this article, we will delve into the current state of AI and its anticipated developments in 2024.
Transition from Classical AI to Generative AI
The inception of OpenAI’s large language model (LLM)-powered chatbot,
ChatGPT, marked a significant milestone in the evolution of AI. GenAI systems such as Google’s Gemini 1.0, Meta’s LLaMA 2, Anthropic’s Claude 2, and others have further propelled the capabilities of AI beyond traditional machine learning. These systems have the unprecedented ability to instantly generate diverse forms of content, including articles, images, videos, and presentations, based on natural language prompts. Moreover, the emergence of smaller language models (SLMs) has made domain-specific tasks more accessible and energy-efficient.
The implications of GenAI are profound, with research firm Gartner predicting its substantial impact across various sectors such as pharmaceuticals, manufacturing, media, architecture, and medical sciences. These advancements are poised to significantly contribute to India’s GDP and revolutionize marketing messages and drug discovery by 2025. However, enterprises are exercising caution due to the potential misuse of GenAI for creating fraudulent content and violating data protection regulations.
Ring-fencing AI
The exponential growth of GenAI models has raised concerns regarding the development of artificial general intelligence (AGI) or artificial super intelligence (ASI), which could mimic human cognition, potentially replacing numerous jobs and posing security risks. This apprehension is shared by industry leaders like
Elon Musk and Masayoshi Son, who envision ASI surpassing human intelligence by a staggering factor. Conversely, other AI experts emphasize the myriad benefits of AI, such as its application in smartphones, autonomous vehicles, satellites, and disaster management systems. Nonetheless, there is a unanimous consensus that AI and GenAI need to be regulated, leading to the formulation of comprehensive AI Acts and regulations by various governments worldwide.
For instance, the European Union (EU) has proposed a risk-based AI Act to address biometric categorization systems and unauthorized scraping of facial images, aiming to mitigate privacy violations. Similarly, the Group of Seven (G7) nations and India have initiated efforts to establish guiding principles and regulations for AI. India, in particular, is on the brink of implementing the Digital India Act (DIA) to regulate AI and intermediaries while preserving a balance between leveraging AI’s potential and curbing its misuse.
The Indian LLMs
In a move tailored to meet the specific needs of the country, India has embarked on developing its own large language models (LLMs). Akash Ambani, the chairman of Reliance Jio Infocomm, announced a collaboration with IIT Bombay to create ‘Bharat GPT’, an LLM customized for India. This initiative reflects the endeavor to harness AI to address India’s unique societal and business requirements.
In recent years, there has been a significant
push towards developing India-specific Language Models (LLMs) to cater to the linguistic diversity of the country. The announcement of ‘Bharat GPT’, an LLM tailored to suit India’s needs by Reliance Jio Infocomm, in partnership with the Indian Institute of Technology (IIT) Bombay, reflects this growing trend. Similarly, Krutrim, a new company founded by Ola co-founder Bhavish Aggarwal, is also focusing on building India-specific LLMs. However, the task is not without its challenges, given the high computing costs and the limited availability of good Indian data sets.
The Linguistic Diversity of India and the Data Challenge
India is home to over 400 languages, making it one of the most linguistically diverse countries globally. Most foundation models and LLMs are trained using predominantly English internet data, posing a challenge for incorporating the richness of Indian languages. Many of the 22 official Indian languages lack digital data, making it difficult to build and train AI models with local data sets. Efforts such as those by Bhashini, a unit of the National Language Translation Mission, have been instrumental in collecting and curating data from different sources to address this gap. This has involved a significant investment and a large workforce to collect and annotate data, both in text and speech, for the development of India-specific LLMs.
Initiatives and Collaborations for India-Specific LLMs
Several initiatives and partnerships have emerged to address the need for India-specific LLMs. The ‘Nilekani Center at AI4Bharat’ at IIT Madras is focused on building open-source language AI for Indian languages, encompassing data sets, models, and applications. Additionally, startups like Sarvam AI and CoRover, in collaboration with organizations like Bhashini, are developing indigenous LLMs tailored to serve diverse Indian languages. The Mahindra Group also aims to launch an indigenous LLM, known as the Indus Project, targeting the inclusion of multiple Hindi dialects. These collaborative efforts underscore the industry’s recognition of the importance of catering to India’s linguistic diversity through LLMs.
Unveiling the Potential of Quantum Computing
Quantum computing is poised to revolutionize the way information is processed and holds tremendous potential for a wide array of applications. Unlike traditional computers that rely on bits, quantum computers utilize qubits, which can process information in multiple states simultaneously due to a property called superposition. This exponential growth in computing power has far-reaching implications, especially in fields such as molecular structure simulation, drug discovery, and complex data analysis. Companies and research institutes are actively engaged in developing ‘fault-tolerant’ quantum computers, with the aim of enabling real-world applications across various domains.
AI's Impressive Advancements and Energy Challenges
The advancements in AI have paved the way for impactful applications across diverse sectors, including disaster management and flood forecasting. AI-enabled systems have the potential to mitigate the impact of natural disasters by aiding in the design of better warning systems and the deployment of smart robots and drones for disaster management. However, it is crucial to acknowledge the energy-intensive nature of AI and generative AI tasks. The generation of new content, such as text, summarization, and image generation, incurs significant energy consumption, and the development of GenAI models further amplifies this trend.
Concluding Remarks
The endeavor to develop India-specific Language Models reflects a significant step towards inclusive technological innovations. The concerted efforts by industry players, academic institutions, and startups to address the challenges associated with linguistic diversity and data availability are commendable. Similarly, the exploration of quantum computing’s potential and the application of AI in disaster management underscore the transformative impact of these technologies. However, the energy implications of AI and generative AI call for a balanced approach, emphasizing the need for climate-friendly data centers and IT infrastructure as a sustainable pathway forward. As India continues to embrace technological advancements, the development of India-specific LLMs and the responsible deployment of emerging technologies will play a pivotal role in shaping the nation’s digital landscape.
