CHATGPT : - AN AI TOOL

Research workers of MIT University created a computer programme in 1964, it provided seamless communication between human & machines. By 1966, they built an app called ELIZA, that helped in building the chatbots.  Eliza was equipped with, the  keyword-enabled responses. 

Another language processing bot, A.L.I.C.E, was introduced in 1995 & in 2001 a new version ‘SmarterChild’ was launched. It was considered as the foundation of modern AI chatbots. A decade later, the introduction of AI chatbots like, Amazon Alexa, Google Now and Siri, were introduced. 

An AI powered chatbot was founded in 2015 by a group of entrepreneurs, including Elon Musk, Sam Altman & Greg Brockman. An artificial intelligence research firm, called OpenAI introduced an advanced version of AI powered chatbot, on November 2022. 

ChatGPT is much more than a chatbot,it is called as, a ‘Language Model’. User can ask it to write a program or for a creative task such as writing a story. It is good at explaining scientific & factual concepts. A language model is a software that can express a sequence of words as an output with appropriate semantic relation & a basis of  input commands. So, it enables the system to perform creative tasks like answering questions & carrying on a conversations. It is used in ‘natural language processing’ (NLP) applications, like speech recognition, automatic translation. 

It functions, as a ‘neural network’, which is a combination of a large network of computers that, collectively derives output of words based on the, commands given to it, during its training stages. Such training process of a neural network is called ‘Reinforcement Learning’. The input data is typically huge corpus of text. All these technologies are part of the artificial intelligence, which is also called ‘Machine Learning’. 

It has been showing tremendous advancements. A very important part of the technology is, “word embedding” which represents words as a matrix of numbers that can be manipulated inside computers. A neural network analyses these numbers, it can interpret the different contexts of numbers. A neural network has the accuracy in analysing the exact context of a sentence, when it is supported by a refining technique called “Transformer”. Such comprehended data is used for performing multiple tasks like answering a question, summarising a paragraph. 

Such language models were released by OpenAI from 2018, which are part of ChatGPT.  They are popularly called as “Generative Pre-Training (GPT) language model”. This technology makes it suitable for creative tasks like writing a new story. Such type of neural network can generate new content, based on the input content, called training data. This technology can generate creative output like fictional stories. 

A new technology called, “Generative Pre-trained Transformer 2”(GPT-2) was developed by upgrading the ‘transformer’ technique, in 2019. Till the year 2022, its upgraded version, GTP 3.5 was launched. Now ChatGTP, is considered as a successor of GTP 3.5. 

ChatGPT is empowered with communication skills which makes it smart in conversational responses. This accuracy is the result of, additionally trained neural network ,on human feedback with conversational transcripts.There are alternative language models in the AI community, like BERT (Bidirectional Encoder Representations from Transformers) from Google. However, ChatGPT is considered  most powerful for communicative functions. 

THE CONVERSATIONAL AI 

The functional accuracy of  any language model can be measured using a technique, called as “Recall-Oriented Understudy for Gisting Evaluation” (ROUGE). 

Another metric that is used to analyse the functional accuracy of language models is called the BLEU metric (Bilingual Evaluation Understudy); it compares overlap in translated content with a standard translation.

ChatGPT can become an effective educational tool for any topic and for all age groups, as it can be even instructed  to explain any information to a kid. It has adopted to the art of easy explanation in simple terms . 

However, it may not be equally skillful in specialised contexts. For such assistance the neural networks behind ChatGPT need to be trained on specialised data.  As ChatGPT can write programs, it may be  possible to make it knowledgeable on any specialised topic eventually. 

ChatGPT has cleared an MBA examination set by Wharton Business School faculty, and performed well in law examinations. It wrote a piece of law whose objective is to regulate the use of such chatbots. It is useful in writing essays or in drafting the petitions. 

PROBLEMS IN USING CHATGPT

The functions of ChatGPT, can influence, the role of teachers in interactive learning like oral examinations & in-class proctored exams  & interactive lectures. The AI system & its users may evolve sophisticated means & ways to  excel in all these areas . The graph of users showing rapid hike, may create situations, to assign a monitor or a guide.  

In today’s times, the knowledge on diversified issues is available at fingertips. Even material of specialised level is available at different sources, but there are very few, who visits such a deep level of content. A trend of decreasing attention spans & general distractedness is contributing, for selective reading and studies in today’s times. 

ChatGPT & other such technologies, can enhance such tendencies. Moreover its use may ensure the outsourcing of thinking and analysis. 

There is a limitation in the use of ChatGPT, is that, the quality of the output content is dependent on the inputs given in the form of trainings. It means the quality of the training model will determine the output quality of the application. One more limitation is about the, alternate facts and views as well as analysis, even if they are half-truths or unauthentic information, they too will be the part of the output, without its user being able to detect it.

The New York City department of education, banned the use of ChatGPT in public schools & some universities. The Italian data protection authority ‘Garante’ ordered the company to temporarily stop processing Italian user’s personal information, while it investigated a possible data breach. 

Some supporters are associating the, ChatGPT & similar AI systems with, equality in terms of access to knowledge. But in such a scenario, the need for well-qualified teachers increases, who can keep a check on the use of technology and can ensure the positive use of such an advanced technology.

According to Bloomberg, Microsoft would invest $10 billion over several years in OpenAI after the $ 1 bn it had invested in 2019 & a round more in 2021. Such a huge investment is not done unless, it promises a huge change in the market or in terms of promising returns it makes. 

DISAPPROVING THE LINGUISTIC DIVERSITY

When a dominant social group, patronises a mode of writing, it becomes dominant. Such patronisation involves a strong assertion over other social groups. There are different forms of English and the ‘standard American English’ is prioritised  in many American classrooms. On the other hand, the speakers of Black English, a cultural language with its own grammar & a remarkable history of resistance are penalised & shamed for its difference in forms and patterns. 

Similarly, in Aotearoa New Zealand, the Queen’s English became dominant not because it’s intrinsically better than te reo Māori. Rather, European colonisers wanted to stamp out Māori culture, and writing in the Queen’s English became a key tool for furthering that objective.

According to supporters of ChatGPT, it can read, analyse and generate content in many languages, including in Black English and te reo Māori. But the controversy is about its default settings. ChatGPT is configured to treat some forms of writing as normal, typical & expected. ChatGPT requires a special request to generate non-normative forms of writing.

Similar default behaviour also occurs in it’s sister programme, Dall-E 2 . When this AI neural network shown the image of ‘typing hands on a laptop’, It displayed four images & all were ‘white masculine hands’. Such kind of  algorithmic bias continued to portray white English-speaking men as a standard of writing culture, which is assuming and pushing others to the periphery. White English-speaking men have dominated literary  sectors, including journalism, law, politics, medicine, computer science and academia.