WITHcompany OpenAI now allows users to create their own versions of the chatbot ChatGPT, which the company calls GPT for short, for personal and professional tasks. They can, for example, teach a child math or book a flight. You don’t need to know how to program to create them. The company announced this at its first developer conference. Company CEO Sam Altman also introduced a trial version of the GPT-4 Turbo chatbot, a faster and more powerful version of the company’s Large Language Model (LLM). The head of Microsoft, Satya Nadella, also spoke briefly at the conference.
The OpenAI start-up, which launched the ChatGPT tool last November, enjoys the support of Microsoft, which invested ten billion dollars (2.3 trillion crowns) in the company. ChatGPT currently has around one hundred million users each week, Altman said. Over 90 percent of US Fortune 500 companies build tools on the OpenAI platform. More than two million developers use OpenAI technology to create generative applications.
This month, OpenAI plans to launch a store where users can find GPT chatbots customized by other users, or earn money through their creations, similar to Apple’s App Store. Altman said that in the future, he envisions everyone having multiple GPT chatbots that work together and complete tasks for them.
Examples of such GPT tools that the company has published on the Internet include Laundry Buddy, i.e. a laundry helper. The user can “ask everything about stains, washing machine settings and sorting laundry”.
The GPT-4 Turbo version relies on Internet data up to April this year, so it has a better overview of current events. The GTP-4 version had access to data until September 2021. However, the company introduced a feature this year that allowed ChatGPT users to browse the Internet and get up-to-date information.
“We were as bitter as all of you, and perhaps even more, that GPT-4’s knowledge of the world ends in 2021,” Altman said at the conference. “We will try to make it never so out of date again,” he added.
The Turbo version of ChatGPT will be able to process and respond to user input, which will be as long as a novel. The company’s GPT-4 model, by comparison, was limited to text up to about 50 pages. The Turbo version will also be cheaper, according to OpenAI.
OpenAI has also released a series of updates aimed at developers. These, together with a significant reduction in costs, met with an enthusiastic response from the audience, Reuters wrote. “This is a huge benefit for start-ups like us. Our costs suddenly decreased three times,” said Flo Crivello, founder of the Lindy start-up. On the other hand, he admitted that his company could compete with new GPT tools developed by users in the future.
The intention of OpenAI, according to Reuters, is for more businesses and developers to create models to compete with those from developers at companies such as Anthropic and Google, and open-source models such as Meta Platforms’ Llama.