By Rishi Kant, Joint Director, Directorate of Economics and Statistics (DES)
ChatGPT, chatbot developed by OpenAI, has taken the world by storm. Within two months after launch, it is estimated to have more than 100 million active users, clearly reflective of its massive popularity. Chatbot GPT an AI large-scale language model (LLM), is the latest manifestation of AI powered by vast quantities of data, algorithms capable of correcting themselves, and massive computational power. It uses a deep learning architecture which allows it to understand and generate human-like text and responses, by using mix of vast datasets and powerful computing. It can take a user’s input (a prompt or question) and generate a relevant and coherent response in natural language.
Disruptive technology
As an AI language model, ChatGPT can be considered disruptive as it demonstrates remarkable natural language understanding and generation capabilities. It can comprehend context, follow conversational threads, and provide relevant responses, which has the potential to revolutionize human-computer interactions and can automate tasks that previously required human intervention. AI models like ChatGPT can continually learn and adapt from new data. As more users interact with the system, it gains insights and improves its responses over time, leading to a constantly evolving and self-improving AI.
Potential for Misuse
While ChatGPT has numerous positive applications, it also has the potential for misuse. Besides, concerns about data breaches and surveillance, the use of AI has also raised concerns about its potentially harmful effects on the accuracy and integrity of the information it produces. ChatGPT can be manipulated for generating misinformation, as the model is trained on a large volume of data, and finding associations and patterns without making counterfactual inferences in terms of good or bad. Therefore, risk of spreading misinformation, hate speech, and deepfake content can become much easier, while tackling the same become even more difficult.
Another risk is the potential for generating and perpetuating bias. Such bias can be either in relation to certain religions, gender, race, ideology, caste etc. AI systems learn from data, without making any ethical or moral judgments or reasoning. AI lacks emotional intelligence, empathy, and compassion, which are crucial components of moral decision-making and rest only with the conscious human mind. Devoid of causal reasoning AI system, if trained using data, which advertently or inadvertently contain biases, may perpetuate and even amplify those biases. For example, if historical data contains discriminatory patterns or reflects societal prejudices, the AI system may make biased outcomes. Referring to such outcomes as holy grail may lead to inaccurate or unfair conclusions, which may prove disastrous, especially when such biases creep into the policy-making domain.
Besides, teaching AI to make ethical and moral decisions presents its own set of challenges. The ethical rules, principles, and priorities can vary across cultures, societies and individuals, making it difficult to program a universally “ethical” AI. What is considered ethical in one context may be seen as immoral in another. Moreover, societal values and moral norms can change over time. What is considered ethical today may not be so in the future. Keeping AI aligned with evolving ethical standards requires continuous adaptation.
In certain situations, AI may face ethical dilemmas where no option is entirely ethical, and it has to make trade-offs. For instance, what may sound ethical or moral from the utilitarian perspective may not be so from a deontological perspective. A utilitarian might argue that sometimes it’s necessary to temporarily set aside certain rules in order to achieve the greatest overall well-being, while a deontologist might contend that certain principles are inviolable, regardless of potential positive outcomes.
Overall, ChatGPT and similar AI language models have the potential to bring about significant changes in how we interact with technology and each other. While this disruption holds great promise for innovation and efficiency, it also demands careful consideration of the ethical and social implications to ensure responsible and beneficial use.
