Summary:
Global institutions like the World Bank and the IMF recognize India’s economy as the fastest-growing among major nations. Our study indicates that India could experience a substantial boost in its GDP over seven years (2023-24 to 2029-30). The cumulative impact on GDP may range from US$1.2 trillion to US$1.5 trillion, contributing an additional 0.9% to 1.1% in annual CAGR. Given the immense capability of Gen AI with respect to its productivity and efficiency enhancing effects, its adoption has the potential to accelerate India’s growth trajectory. It is, therefore, necessary to increase investment in Gen AI, education and upskilling to fully capitalize on the demographic dividend.
In line with the development of Digital Public Infrastructure such as the India Stack, Aadhaar, UPI, etc., the government can consider developing Gen AI systems as Public Goods. This approach is in line with the National Strategy for Artificial Intelligence (2018), which emphasizes an inclusive ’AI for ALL” lens.
The development of Gen AI as Public Goods can be a game changer as it can also be deployed across various sectors of impact such as education, healthcare, agriculture, smart cities, etc., where the government is a key player.
Impact of Gen AI on Different Sectors
To promote the development of Gen AI, policy actions will have to ensure access to data. The development of indigenous training data sets (especially for local Indian languages) will be very important. The government may invest in creation of structured and unstructured datasets, which can be opened to the public. These datasets, compiled through various government surveys or generated through administrative processes, can be made available in a format that is easily usable for Gen AI development.
Future Steps:
Marc Andreessen, co-creator of Mosaic browser and now General Partner of Andreessen Horowitz, one of Silicon Valley’s eminent Venture Capital firms, puts it in his essay ‘Why AI will save the world’, “Anything that people do with their natural intelligence today can be done much better with AI.
Code generation becomes the first killer app for foundation models Generating code is arguably one of the early killer apps for Gen AI. Code has an inherent structure that foundation models find easy to learn and predict.
GitHub copilot from Microsoft and CodeLlama from Meta has, in the space of the past few months, become standard tools used by developers. While foundation models can improve productivity across a wide range of repetitive programming tasks, they are also being used to document legacy code, refactoring legacy apps and help enforce coding standards.
Big Tech takes AI into the Enterprise:
AI is no longer a preserve of large companies with terabytes of data and access to supercomputers. Big Tech is incorporating Gen AI into all their offerings and bringing AI enabled capabilities to enterprises of all sizes via their cloud offerings. Google, Microsoft and Amazon have all significantly revamped their cloud services with offerings to enable building AI apps.
- Google is aiming to make Bard the personal assistant on your phone, handling tasks like planning a trip, sending texts, searching messages, etc. and is available through your workspace.
- Pixel phones with AI allow image editing to completely alter entire photos.
- Meta is adding Gen AI into its ad platform to help brands manage their creative campaigns – from editing images of products to writing copy.
- Microsoft is adding co-pilots into all its offerings – Bing, Edge, Windows and Azure. This allows image editing in all photo apps, chat with documents, assistants across all assets, and write emails like you.
- Google has Duet – an LLM-powered chatbot inside Google Cloud. Amazon Bedrock helps companies fine tune their own models. It has invested in Anthropic and made its powerful LLMs, including Claude and Claude 2, available on AWS.
Equipping AI Workforce:
In the near term, businesses grapple with a shortage of individuals possessing AI skills, a challenge that is expected to continue. Despite India’s commendable standing in AI skill penetration and talent concentration, the advent of Gen AI amplifies this shortage. The widening gap between the skills demanded by companies and the existing workforce underscores the urgent need for strategic talent acquisition, particularly for the successful initiation and scalable implementation of prioritized use cases.
Shielding with Responsible AI:
Enterprises must be vigilant in managing Gen AIrelated risks to avoid any reputation or financial loss. C-suite executives and leaders need to proactively comprehend and integrate processes for risk mitigation and governance. At present, organizations see data privacy as the most important risk of Gen AI. To address the concerns posed by Gen AI, enterprises can focus on addressing the following key risks:
1. Trust and Performance Risk: Hallucinations in FMs and LLMs lead to erroneous responses and erode user trust. Enhancing performance involves maturation of LLMs, integration of enterprise data, and employing techniques such as Data Grounding, Dynamic Embeddings, and Reinforcement Learning from Human Feedback (RLHF). Transparency in Gen AI responses, such as sharing data sources and clarifying AI generated content, fosters trust.
2. Bias and Toxicity Risk: Bias in training data and models can lead to unfair outcomes and discrimination. Monitoring, identifying and removing biases through Bias Auditing and Data Fairness tools are crucial. Implementing measures such as meta prompts and content filtering, must be implemented to manage toxic prompts and responses.
3. Security and Privacy Risk: Managing the risk of leaking proprietary and sensitive data to LLMs is a priority. Employee training and technical guardrails are essential to prevent the entry of proprietary data into non-enterprise Gen AI tools. Enhanced cybersecurity measures are necessary to counter external threats such as prompt injections and model theft. Compliance with GDPR (General Data Protection Regulation) and local data regulations is imperative which includes getting user permissions to use their prompt data, transparency on responses generated by AI and data sources used.
4. Regulatory, Compliance and Copyright Risks: Enterprises must stay informed about AI governance, ethics policies, and regulatory provisions. Compliance with evolving regulatory frameworks is essential. Awareness of copyright risks, especially in image and code domains, is crucial. Understanding existing lawsuits against FM companies is also necessary.
5. Ethical Risks: Enterprises must navigate ethical concerns related to job loss, technology misuse (for example, deep fakes), risks of super intelligence, and sustainability challenges. Establishing a clear AI governance framework, AI Ethics Board, and responsible AI practices is crucial for addressing ethical queries from internal and external stakeholders.
The Economic Opportunity of Gen AI in India
India has the potential to add
US$359 billion to US$438 billion to its GDP
on account of Gen AI adoption in 2029-30 over and above its baseline estimates
This represents an additional
5.9% to 7.2% of GDP in 2029-30
Over a period of seven years (2023-24 to 2029-30), Gen AI’s contribution would translate to
US$1.2 trillion to US$1.5 trillion
cumulated GDP impact
Achieving this potential would provide the Indian economy with an additional CAGR of 0.9% to 1.1%
source: AIdea report by EY