There has been a recent faze on the internet around ChatGPT especially among the tech community. Previously, the worlds interaction with AI chatbots has been digital assistants and company bots that people interact with primitively with only an ability to respond to rudimentary questions on help desk pages or Frequently Asked Questions(FAQS). However, the conversation around ChatGPT on twitter proved that this a new phase in the AI world.
So what exactly is ChatGPT?
In simple terms, ChatGPT is an AI chatbot that you can write to and ask questions as though you are speaking to another human and it will respond. It took 5days to cross 1million users, a record breaker for a lot of new products. It was launched end November by OpenAI, same company that did InstructGPT. It uses Natural Language Processing Model and was trained on an Azure AI supercomputing infrastructure. Many are using it as a virtual assistant. How this plays out is what you would normally spend minutes researching on google, ChatGPT will give it to you in seconds. Developers are looking to use it to not only build their code but also optimize it.
ChatGPT answers follow up questions, solves math problems, writes research papers, challenges incorrect premises, generates code and even admits to mistakes amongst other features. Since it was trained on enormous amount of text data, it works on predicting the next text.
Being that it was created by humans, it has some few limitations. Some of its critics have paused its bias nature which is true for many AI models.
It has also been noted to give plausible-sounding but incorrect answers since during its training there was no standard model of truth given. Another limitation is that its very sensitive to tweaks in phrasing. It can fail to give an answer to one thing but one slight change and the model will answer correctly. With its ability to write academic papers and the like, educators are raising concerns around its potential to transform academia, for better or worse, with rising cases in plagiarism.
ChatGPT has been received with both positive and negative reviews. Elon recently tweeted, “ChatGPT is crazily good.” One Guardian piece was quoted saying professors, programmers and journalists could all be out of a job in just a few years. Krishna Gade, CEO of AI model monitoring startup Fiddler foresees ChatGPT and other related language technologies could cause a disruption in web search, a field widely dominated by Google currently. Its growing popularity shows that a large population prefers the question and answer model as compared to search query. However, Google still has a large advantage over ChatGPT because of some of the challenges cited above, with bias being a major concern.
Kelsey Piper of Vox wrote that “ChatGPT is the general public’s first hands-on introduction to how powerful modern AI has gotten, and as a result, many of us are stunned.” It will be interesting to continue exploring the power of AI with the lessons that will be picked from this deployment to future AI models especially in the NLP field.