ChatGPT: Capabilities and Limitations

ChatGPT is a computer program that can carry out an intelligent conversation with a human and answer questions, and generate text based on input. It can answer follow-up questions, admit its mistakes, challenge incorrect premises, reject inappropriate requests, correct grammar, summarize difficult texts into simple concepts, convert movies into emojis, and fix bugs in code, to list a few applications.

Where does ChatGPT come from?

It is based on a form of Artificial Intelligence (AI) called Deep Learning to generate text based on a given prompt. AI has been around for almost 80 years and conversational AI has been considered the most difficult branch of AI due to imprecision, nuances, and context boundedness of natural
languages. This is at variance with branches like mathematics which are characterized by precision. However, ChatGPT was able to defy the odds and carry on a conversation better than humans in some topics.

ChatGPT, released on November 30, 2022, in San Francisco, can be thought of as a virtual writing assistant or a smart chatbot that can understand and respond to humans in natural language. It was developed by OpenAI based on natural language processing which has been trained on massive amounts of text data from various sources such as books, theses, and case studies allowing it to generate human-like responses to a variety of questions and prompts. It is a machine that has read and learned a lot so it can write its replies and stories.

The capabilities of Generative AI (Chatbot)

Unfortunately, Generative AIs can be manipulated to be sources of false information. However, an Application Programming Interface (API) called Moderation API has been developed to counter this. This is a moderation system intended to assist developers when things go against their content policy like illegal or unsafe information. Microsoft and Meta both launched their Generative AIs called Tay and BlenderBot 3 before ChatGPT.

Twitter users, unfortunately, taught Tay rude, racist, and misogynist language which led to its downfall. This was also the case with BlenderBot 3. The Moderation API is not a hundred per cent foolproof but certainly helps.

The limitations of ChatGPT

Despite its capabilities, ChatGPT does have limitations. It may occasionally generate incorrect information, produce harmful instructions or biased content, and has limited knowledge of the world and events post-2021.

There are also ethical implications that come with Generative AIs such as ChatGPT including having programmers using it to write their codes and school kids using it to answer their exam questions. People can also be the target of Generative AI through deep fake videos, explicit content, or propaganda.

Notwithstanding its limitations and ethical implications, ChatGPT is seen as the next era-defining innovation which is bound to change how we interact with the world and the internet as we have come to know it.