A new frenzy captured the tech world as buzz about the new CHATGPT emerged last week, so much so that even the local newspapers ran an article about it. In a groundbreaking innovation, OPENAI, an artificial intelligence company, designed an automated chat programme that could generate human-like responses. It is so emotive that users could get confused and believe that they are indeed interacting with a live human being on the other side. The AI can compose articles, poems, and scripts and summarise texts. Besides, it can fix code issues and answer queries across a wide range of topics. While AI-powered chatbots are not new, this one is seemingly different. Around a million people put it to the test in its initial open trial. One of the most intriguing features is that it admits its mistakes! For some, ChatGPT is forward-looking journalism and education.
OpenAI, co-founded by Elon Musk, is based out of California and intended the advanced ChatGPT to respond to requests in dialogues.
What exactly is ChatGPT?
Machine learning is used to generate text and communication by the GPT-3.5 language model, which seems eerily human. A natural language processing tool, ChatGPT, powered by AI, answers questions and helps you write emails, articles, and code. Even when asked about the meaning of life, the bot could provide a (too) comprehensive explanation. Those who want to experience this new power of AI can always reach out to Twitter to find many mentions.
How does it work?
The OpenAI blog post mentions that the model was first trained using RLHF or Reinforcement Learning from Human Feedback, which uses methods similar to InstructGPT, the only difference being the data collection system. Of course, human AI trainers assayed both roles, user and assistant, to fine-tune the trained model and arrive at this slightly advanced setup. Responses were composed based on model-written recommendations accessed by the trainers. Furthermore, comparison data was also collected consisting of two or more model responses, which were then evaluated for quality.
Potential Issues?
Although the company announced its breakthrough, the technology is still nascent. The chatbot might sometimes enumerate too many details making it excessively verbose. Certain phrases are overused (it keeps stating that OpenAI has created it).