Topper Copies

Generative AI holds the potential to push boundaries of innovation in the near future, yet its societal impact raises substantial challenges. Elaborate.

Kush Motwani, AIR 11, 2023

Generative AI refers to a type of Machine Learning - that can create new and original content (text/images/videos, etc), using certain input prompts, training data, and machine learning algorithms.

It holds potential to push boundaries of innovation in near future as 1. Potential to lead to a new generation of art: DALL-E, etc for image creation; Chat-GPT for text (generative pre-trained transformer) 2. Composition of poetry, music, etc. 3. Innovation in medicine: Generative AI assisted drug discovery, prediction of protein structure. 4. Innovation in design and architecture: Generative AI can make new architectural designs, its simulations can be used to push boundaries of what construction is possible using given materials, and similarly to innovative interior designing 5. Chat-bots: e.g based on Chat-GPT. 6. Remuneration of professional knowledge: - Legal advice; e.g recently a judge of Punjab and Haryana High Court used Chat-GPT for research. - "No-code" coding: e.g GitHub Copilot can help people to code, who don't know how to code, etc.

Yet, its societal impact raises several challenges 1. Governance challenges: pvt entities are themselves governing generative-AI, raises societal concerns around sovereignty 2. Deep fakes: using GANs (Generative Adversarial Networks) - potential to destabilize society with fake news, violate dignity of women (deep fake pornography) 3. Lack of transparency: black box algorithms - even the developers themselves cannot explain their algorithms 4. Fear of job losses: across professions (e.g law, architecture, IT services, etc) -> societal instability (by Sam Altman - founder of Chat-GPT) 5. Reduced development of human mind: if all work is outsourced to AI -> evolutionary threat 6. Intellectual property and copyright issues: if original work is based on data by copyrighted work, then the ethical question is that of allocation/violation of intellectual property 7. National security risks -> information warfare, etc 8. Bias in Generative AI - due to bias in training data.

Way forward 1. Focus on explainability of AI 2. Deep fake prevention and detection (e.g using Reality Defender, DipTrace) 3. Awareness and education 4. Diverse training data - which is unbiased.