Followed by DALL-E 2, OpenAI revealed another remarkable AI tool, ChatGPT, an AI chatbot prototype capable of understanding natural language and responding in natural language. ChatGPT generates human-like responses to text input using advanced machine learning algorithms.
The developers created the AI chatbot on the Azure AI supercomputer based on the GPT-3.5 language model from OpenAI. The fundamental goal of the creators of the chatbot was to make it easy to use, correct and “human”.
ChatGPT was trained using a huge amount of text from the internet and the Reinforcement Learning from Human Feedback (RLHF) system. This means that special AI trainers checked the answers of the model and communicated with it.
The creators warn that the bot works with limitations and bugs. For example, the service is too verbose and may generate plausible but no true answers. It is possible that the user will have to reformulate the message several times in order for the chat to give the correct answer. The developers also taught the AI chatbot to reject “dangerous” questions.
What can ChatGPT do?
As per the current experiments, the OpenAI AI chatbot is capable of writing essays and poetry. The scary thing is it can put programmers out of work because ChatGPT learned how to edit software and write code from scratch like GitHub Copilot, even see software vulnerabilities and suggest ways to fix them.
Here are some of the wonders that ChatGPT can do:
ChatGPT is free to use; all you need to do is register on the OpenAI website and start a conversation. According to OpenAI CEO Sam Altman, the AI tool has already crossed 1 million users.
ChatGPT delights and scares web users at the same time. Commentators worry that in the future, neural networks may be able to do most of the intellectual work for a person.