The cyber world was taken by storm with the launch of ChatGPT on November 30, 2022. Developed by OpenAI, this chatbot powered by an Artificial Intelligence engine provides a significant boost to the technological and knowledge domain in the 21st century. Chatbots are software programmess that utilize artificial intelligence and natural language processing to emulate human conversation through text, voice or both. ChatGPT has the ability to answer a wide range of questions, much like a human being, and does so by mining various data sources. The launch of ChatGPT became a significant milestone in the development of AI-powered solutions for knowledge and information-based industries. It revolutionized the way we approach problem-solving, research and decision-making in the 21st century.
ChatGPT can be assigned roles to customize the responses. The type of answer provided by ChatGPT will depend on the role assigned to it. If you ask a general question, then it will provide a general answer. For instance, if you ask "Describe the culture of Manipur" it will give you a general answer about the culture of Manipur. The text you got may be appropriate for a postgraduate holder but not for class six students. But, if you assign a specific role to ChatGPT, it can provide a customized and more specific response.
If you ask "Consider yourself as a school teacher. How will you describe the culture of Manipur to the students of class six standards", it will provide a presentation that is more appropriate and understandable to the class six students. This feature is an important functionality of this chatbot. It means that the chatbot can emulate the personality of the role assigned and customize its responses to the intended audience. It would be significant when we used it for educational or customer service sectors where tone and style of communication are important.
The ChatGPT can emulate human thought processes and creativity to some extent. If you asked this chatbot to write a new poem describing Manipur University, it can generate a new poem in different versions. The user can regenerate the response and compare the outcomes to select the best one. The user can also instruct the chatbot to write a new poem in another language and it will produce the output in that language. Many languages are supported in the chatbot, including Hindi, Bengali, etc.
The aesthetic quality of the generated poem is a matter of subjectivity, the stark reality is that a machine can produce a new poem and is a remarkable feat. It demonstrates the potential of AI-powered tools in creative fields including writing and art. It opens a new vista for expanding the boundaries of what is possible.
When I use the chatbot in the context of Meiteilon (Manipuri) language, it gives some unwanted responses. I instructed the chatbot to write a new poem in Meiteilon describing Loktak Lake and to represent the text in Meetei Script. As expected it provided the output in Meitei Script (Meitei Mayek). Surprisingly, the language is in the Bengali language. Why it is so?
Meiteilon and Bengali are distinct languages with their own unique vocabulary, grammar and syntax. This error might be stemmed from limitations in the language identification process. In the chatbot model, the AI-program specially trained on input texts available and generated the response.
The ChatGPT might not have trained specifically in Meiteilon but more trained in the Bengali language. This chatbot is based on GPT-3.5 (Generative Pre-trained Transformer) architecture which is a deep neural network. It is trained on a large corpus of text data but does not possess the full range of cognitive abilities.
Current limitations are not necessarily permanent as ongoing R&D is trying to improve the capabilities. On the other hand, hundreds of AI-powered apps using various models and techniques have been released after the launch of ChatGPT with similar and parallel features.
In response to the swift advancement of AI technology fueled by ChatGPT and others, a petition calling for a cessation of the AI race for six months was signed by Elon Musk, Steve Wozniak and numerous scientists. The petition was signed for promoting responsible AI development and ensuring that technology is used ethically and for the benefit of humanity. R&D is continuing.
If an AI-based language model is trained on a large and diverse set of data for that language, it will help break the language barriers across the globe. It will have a profound impact on research, content creation, customer service, education and wide range of benefits to businesses, individuals and society as a whole.
In the present scenario, asking questions in one language and getting answers in another language is possible to some extent, but limited by factors like language pair, accuracy and complexity of the language. But in the near future, such limitations may disappear.
If a particular language is not trained by any language models in the future, the speakers of that language will face consequences in education, research, customer service and many other aspects. Lack of language support could limit access to information and communication and may potentially contribute to further marginalization and inequality.
It is expected that investment in language technology and language models could promote access to information and communication for all. But what about marginalized societies and indigenous people? When OpenAI the firm behind the ChatGPT launched Whisper, a speech recognition tool for audio transcription and translation of the Maori language into English, it produces a great debate. This innovation is expected to help in preserving and reviving the languages.
However, it raises another unwanted dimension. Data mining without the consent of the indigenous people may resort to abuse, distorting the indigenous culture and depriving minorities of their rights. It will be the rise of new colonization in the information world.
The world is changing so fast. Coming to the context of Meiteilon, are we prepared? Or are we going to block ourselves from the changing cyberspace? Can we restrict the scenario ad infinitum.
(The views expressed is personal)