ChatGPT, a prototype dialogue-based artificially intelligent (AI) chatbot, has created a new storm worldwide and has over a million users in just five days.
Just for information, Netflix took nearly 41 months, Facebook about 10 months and Instagram 2.5 months to on-board one million users. ChatGPT is trained to produce intelligent and informative text on any topic. It is the latest chatbot from the OpenAI Foundation, founded by many venture capitalists like Elon Musk and Sam Altman, chief executive officer (CEO) of OpenAI.
Initial user response to ChatGPT is quite impressive and people are finding newer uses of the chatbot. Despite its limitations, many users have found ChatGPT very useful in certain 'routine' things. For example, it can generate code to build an app or website from scratch when the user does not know coding. On the other hand, some users who know coding are using ChatGPT to debug codes they had written. Some think ChatGPT can be embedded and used to respond to generic emails or messages.
However, ChatGPT is not without issues. ChatGPT works in a question-and-answer module, where the user can ask a question and, using its training in AI and machine learning (ML), it provides answers through a conversational interface. ChatGPT's AI training contains large volumes of text available in the public domain.
It also means the answers given by ChatGPT are not always accurate or reliable. The reason is ChatGPT is an extensive language model trained by OpenAI and does not have the ability to browse the internet or access information outside of what it was trained. For example:
According to Mr Altman, the CEO of OpenAI, "ChatGPT is incredibly limited but good enough at some things to create a misleading impression of greatness."
On its introductory page, ChatGPT clarifies that while safeguards are in place, the system may occasionally generate incorrect or misleading information and produce offensive or biased content. It is not intended to give advice. "These models were trained on vast amounts of data from the internet written by humans, including conversations, so the responses it provides may sound human-like. It is important to keep in mind that this is a direct result of the system's design (i.e. maximising the similarity between outputs and the dataset the models were trained on) and that such outputs may be inaccurate, untruthful, and otherwise misleading at times."

"ChatGPT is not connected to the internet, and it can occasionally produce incorrect answers. It has limited knowledge of the world and events after 2021 and may also occasionally produce harmful instructions or biased content. We would recommend checking whether responses from the model are accurate or not."
ChatGPT is fine-tuned from GPT-3.5, a language model trained to produce text. ChatGPT is optimised for dialogue by using reinforcement learning with human feedback (RLHF) which uses human demonstrations to guide the model toward desired behaviour.
During the initial research preview, ChatGPT is free to use. You need to create a login ID and password to start using ChatGPT. After login, you can start asking any question to the chatbot. Based on the question and its AI and ML training, it provides answers, sometimes too long, to the queries.
For example, I asked about India in the FIFA football World Cup. It told me, "Indian national football team has not had much success in international competitions such as the FIFA World Cup."
Then the chatbot describes some success the Indian football team had in some regional competitions. However, it ends with, "It is possible that India's performance in international competitions such as the FIFA World Cup may improve in the future, but it will likely require continued investment in the development of the sport at the grassroots level."
I also asked ChatGPT about India in 2050. "It is difficult to say with certainty what India will be like in 2050, as a lot can change in the world over the next 30 years. However, based on current trends and developments, it is likely that India will continue to be a major player in the global economy and a key player in the geopolitical landscape."
ChatGPT cannot distinguish whether the information is accurate or misleading. However, as the developers have said, it is still a work-in-progress.
Further, while ChatGPT can remember what the user said earlier in the conversation, there is a limit to how much information it can retain. The current model can reference up to about 3,000 words (or 4,000 tokens) from the current conversation—any information beyond that is not stored.
ChatGPT is being made available as a research preview so we can learn about its strengths and weaknesses.
Next Part: Will ChatGPT be Useful?
- Prof. Dr. J.V. Yakhmi (ex-BARC)