On Tuesday, March 14, 2024, Microsoft baked-up startup OpenAI announced the initial release of GPT-4, a powerful artificial intelligence model that support texts as well as images.
Background
In his tweet, the chief executive of the company, Sam Altman, called the latest model the most capable and oriented to human values and objectives, while still being a fallacious model. In his online demonstration of the new version, Greg Brockman, the president of the startup company, displayed how GPT-4 could assist users in calculating their taxes. The demonstration also showed that the technology could take an image of a manually drawn mock-up of an elementary website as an input and develop a real one based on it.
GPT-4
In a blog post, the developer of chatbot sensation ChatGPT, OpenAI, revealed that software developers with a waitlist and ChatGPT Plus subscribers can access the text-input feature. The image-input feature will remain a preview of the research. While GPT-4 may resemble much of the prior version, GPT-3.5, in terms of casual parley, the former is claimed to be more creative, reliable, and capable of tackling substantially more abstruse instructions.
This latest version is considered to be more improvised and to possess a greater threshold for handling the complexity of the tasks as compared to its previous version. According to OpenAI, the latest model secured a rank among the top 10% of participants in a bar exam, which is essential for a U.S. law school graduate to clear before their professional practice, while the older model secured a rank among the bottom 10%.
ChatGPT: the fastest emerging application in the history
Chat Generative Pre-trained Transformer, abbreviated as ChatGPT, is the buzzing term nowadays. It is a conversational chatbot that generates human-like text and was launched by OpenAI as a prototype on November 30, 2022. Its ability to generate results with the utmost accuracy and detail has been a prime reason for its quick popularity and has thus taken the world by storm. It is a fine-tuned language model with reinforcement and supervised learning techniques that uses an encoder-decoder-based neural network.
This network has been trained on an Azure AI supercomputing infrastructure and uses a massive amount of data using Reinforcement Learning from Human Feedback (RLHF). It can operate complex instructions and be utilised for language tasks like language translation, writing blogs, emails, answering questions, generating ideas, and summarizing text for a list of endless domains. For NLP-dependent businesses, such as e-education, customer service, etc., ChatGPT is a revolutionary tool. The capability of ChatGPT to respond to unseen domains without extensive training is an impressive feature of this conversational AI chatbot and thus distinguishes it from traditional chatbots in Silicon Valley.
Microsoft And OpenAI
The technology was developed by OpenAI, which is a San Francisco-based research company founded in 2015. It is backed by LinkedIn co-founder Reid Hoffman, Khosla Ventures, and the great tech giant, Microsoft, which is planning to invest $10 billion in ChatGPT in the form of cloud computing and cash. The CEO of SpaceX and Tesla, Elon Musk was one of the founders but resigned from the boards in 2018, currently being one of the donors. The critics and analysts are emphasizing that ChatGPT has wreaked havoc on many applications, including Google’s advertisement revenue business.