OpenAI is now letting users build custom versions of ChatGPT to accomplish specific personal and professional tasks as the artificial intelligence start-up works to beat back competition in an increasingly crowded market.
With the new option, users will be able to quickly create their own specialised versions of ChatGPT—simply called GPTs—that can help teach math to a child or explain the rules of a board game, the company said on November 6. No coding is required, the company said. OpenAI also plans to introduce a store later this month where users can find tailored GPTs from other users—and make money from their own—much as they might with apps in Apple Inc.’s App Store.
The announcement came at its first-ever developer conference on November 6. OpenAI also said it is introducing a preview version of GPT-4 Turbo, a more powerful and speedier version of its most recent large language model, the technology that underpins ChatGPT.
Confronting well-funded rivals
ChatGPT was released to the public a year ago in November, kicking off a global frenzy around all things AI. Roughly 100 million people now use ChatGPT each week, the company said at the conference, and more than 90 per cent of Fortune 500 businesses are building tools on OpenAI’s platform. But the ChatGPT maker is also confronting rival products from well-funded AI start-ups, tech giants and, most recently, Elon Musk, an early OpenAI backer.
Also Read | Boost to global tech sector as ChatGPT spurs AI investment bonanza
For OpenAI, the conference represents a chance to show how much influence it wields over the developer community. Hosting a developers conference is also standard for leading tech companies, including Apple, Alphabet Inc.’s Google and Meta Platforms Inc.’s Facebook. Often, these annual events offer a chance for tech companies to preview major software or product updates.
OpenAI chief executive officer Sam Altman delivered a keynote presentation on stage and was joined at one point by Satya Nadella, CEO of OpenAI-backer Microsoft Corp. Nadella touted their partnership—even as the companies compete for business customers.
Sam Altman (left), CEO of OpenAI, appears onstage with Microsoft CEO Satya Nadella at OpenAI DevDay, OpenAI’s first developer conference, on November 6 in San Francisco.
| Photo Credit:
Barbara Ortutey/AP
Altman gave a live demonstration of GPTs, which is rolling out this week to ChatGPT Plus and Enterprise users. Altman typed a summary of what he wanted to create—a chatbot to give advice to start-up founders—and ChatGPT spit out a suggested name, Startup Mentor, and a profile picture. People can configure specific capabilities for their chatbots, he said, and upload files for it to consider as it is interacting with users. Speaking to a group of journalists after his presentation, Altman said, “It’s very silly but that thing I built on stage is something I’ve been waiting like a decade to build.”
OpenAI said the Turbo version of GPT-4 was built with a trove of online data running through April of this year, giving it a greater awareness of current events. The original version of GPT-4 had access to data running through September 2021, though the company rolled out a feature this year that enabled ChatGPT users to browse the internet to get up-to-date information.
“We are just as annoyed as all of you—probably more—that GPT-4’s knowledge about the world ended at 2021,” Altman said on stage at the conference. “We will try to never let it get that out of date again.”
OpenAI said the Turbo version of ChatGPT will be able to process and respond to novel-length prompts from users. By comparison, the company’s GPT-4 model has been limited to as much as about 50 pages worth of text. Turbo will also be cheaper for developers to use, the company said.
Copyright concerns
Founded in 2015, OpenAI has put out numerous AI models over the years. The technology has become more adept at what is known as generative AI—software that can ingest a short written prompt and spit out content in response, whether it is text that can mimic what is written by humans or realistic-looking images.
Also Read | ‘Digital inequalities will power digital colonialism’: Arvind Narayanan
Some people have already used OpenAI’s tools to write lyrics, draft emails, do homework assignments, and create children’s books. But OpenAI and its rivals have also ignited a new wave of copyright concerns. On November 6, OpenAI said it would pay any costs users incur from copyright infringement claims. Microsoft and Google have previously taken similar steps.
OpenAI’s event was held just blocks from San Francisco’s Hayes Valley neighbourhood, which some have nicknamed “Cerebral Valley” for the growing number of AI start-ups based there. The venue, SVN West, is a multi-story event space that in past incarnations was a ballroom and, more recently, a Honda dealership.