Supported by
OpenAI Lets Mom-and-Pop Shops Customize ChatGPT
With a new GPTs service, anyone can build a chatbot and instantly share it with friends, family and the general public.

Reporting from San Francisco
OpenAI said on Monday that it had created a service that allows individuals and small businesses to build customized versions of its popular online chatbot, ChatGPT, and instantly share them on the internet.
Through a new service called GPTs, anyone can quickly customize the chatbot for a particular task without help from additional software or computer code. The owner of a small bed-and-breakfast, for instance, could build a chatbot that answers questions for anyone who stays there.
“You can imagine someone building a GPT that specializes in teaching someone a new language or gives interior design advice,” Peter Deng, OpenAI’s vice president of consumer and enterprise product, said in an interview.
OpenAI, the San Francisco artificial intelligence start-up, has accelerated the release of its A.I. tools in recent weeks. In September, it folded its DALL-E image generator into ChatGPT and released a new version of its popular chatbot that interacts with people using spoken words, much like Apple’s Siri digital assistant.
ChatGPT attracted hundreds of millions of users after it was introduced at the end of last year, wowing people with the way it answered questions, wrote term papers and discussed almost any topic. Several other companies, including Google and the start-up Anthropic, released similar chatbots. Now, OpenAI is trying to push ahead of its rivals.
The new GPTs service is available to anyone who uses ChatGPT Plus, a version of the chatbot that sells for $20 a month.
Millions of seasoned software developers are already using ChatGPT’s underlying technology, GPT-4, to build their own applications, including everything from automated tutors to search engines. The new GPTs service is aimed at a different audience: individuals and small businesses with no experience as software developers.
Anyone can build a customized chatbot by providing a handful of instructions and, in some cases, uploading a few documents. When a bed-and-breakfast owner designs a bot that answers questions for guests, for instance, he or she might upload the B&B’s existing guest manual.
The owner might also ask the service to create a logo for the new bot using OpenAI’s image generator.
Mr. Deng acknowledged that any bot created with the service would be prone to mistakes. Trained on enormous amounts of data culled from across the internet, technologies like ChatGPT sometimes “hallucinate,” or make stuff up.
Because the technology can also be used to generate offensive, untruthful and even dangerous material, OpenAI said it would vet all new bots created with GPTs and disallow any that violated its terms of service. In the coming weeks, it will offer these bots through an app store similar to the one that Apple offers for iPhone applications.
The company also said any users of the service could request that their documents and other data not be used to train future versions of OpenAI’s technology.
OpenAI is facing lawsuits from writers, artists and computer programmers who argue that the company illegally used their work to build its A.I. systems.
The company said that over 100 million people actively used ChatGPT each week, and that over two million software developers used GPT-4 and other services to build apps.
Cade Metz is a technology reporter and the author of “Genius Makers: The Mavericks Who Brought A.I. to Google, Facebook, and The World.” He covers artificial intelligence, driverless cars, robotics, virtual reality and other emerging areas. More about Cade Metz
Explore Our Coverage of Artificial Intelligence
Latest News
Research from Vectara, a new start-up founded by former Google employees, estimates that chatbots invent information at least 3% of the time — and as high as 27%. Experts call this chatbot behavior “hallucination.”
Rishi Sunak, the prime minister of Britain, hosted dozens of government leaders, tech executives and other experts as part of the A.I. Safety Summit he organized to discuss the dangers of the technology.
President Biden signed a far-reaching executive order on A.I. that imposes new rules on companies and directs federal agencies to begin putting guardrails around the technology.
The Age of A.I.
The F.D.A. has approved many new medical products that use A.I., but doctors remain skeptical of these tools.
A nonprofit led by a respected computer scientist is trying to democratize A.I. by building a freely available alternative to those offered by the likes of Google and OpenAI.
OpenAI now lets outside businesses and independent developers tweak what its chatbot does. A new paper says that can lead to trouble.
Advances in A.I. are offering new ways for bad actors to misappropriate online content of children. That is posing new risks for parents posting about their children on social media.
Researchers at Anthropic asked roughly 1,000 Americans to write rules for their chatbot. The results could be a model for future kinds of A.I. governance.
Leading A.I. researchers are transforming chatbots into a new kind of autonomous system called an A.I. agent that can play games, query websites, schedule meetings and build bar charts.
Advertisement

