ChatGPT, developed by AI company OpenAI, kicked off an explosion of interest in artificial intelligence when it launched late last year, spurring Big Tech companies to race to provide competing tools. Computer coders, office workers and students have been using it to speed up their work and ask questions on a range of topics.
In Silicon Valley and beyond, chatbots have been a staple of dinner-party conversations. Some companies even fired copywriters and replaced them with ChatGPT. But the drop in usage suggests that the tech’s limitations are catching up with it and at least some of the hype surrounding chatbots is overblown.
“You had this moment where it was like ‘oh my God it’s awesome,’” said Sachin Dev Duggal, chief executive of Builder.ai, a start-up that uses artificial intelligence to help people build mobile apps. Then, as people began to encounter the chatbot making up false information, they realized it wasn’t as broadly useful as they initially thought, he said.
A spokesperson for OpenAI declined to comment.
ChatGPT rocketed to an estimated 100 million monthly users in its first two months, according to a report by analysts from the Swiss bank UBS. The bot’s ability to have complex conversations, write poetry and pass professional exams impressed regular users and AI experts alike. Tech pundits called it the fastest-growing consumer app in history, and its rapid growth set off an arms race among Big Tech giants to push out their competing products.
Executives from Google and Microsoft have heralded AI as the next revolution in computing that will completely change how people interact with the digital world. Big Tech firms and start-ups are investing billions of dollars in the tech, and companies are reorienting their entire businesses around AI. Regulators around the world are scrambling to understand the tech and pass laws to ensure it isn’t used to harm people.
But over the past several months, the problems with generative chatbots such as ChatGPT have been laid bare. They routinely make up false information and pass it off as true, a problem that Google, OpenAI, Microsoft and other AI leaders have yet to find a solution for. Some users have complained that ChatGPT’s answers have actually gotten worse over time, especially when it comes to generating computer code.
Many companies have also banned their employees from using ChatGPT at work out of concern that putting sensitive company data into the bot may lead to data leaks.
Running AI chatbots takes an enormous amount of expensive computer-processing power, and analysts have theorized that the drop in quality has come because OpenAI is trying to lower the cost of running the bot. The end of the school year in the United States and Europe could also have triggered the drop in usage, as students who had been using it to write papers go on summer break.
Others suggest that concerns over looming regulation and new rules in the European Union have made OpenAI and other AI companies turn down the abilities of their chatbots so as not to run afoul of politicians concerned about the bots being used to spread misinformation, infuse bias into more tech products and impact the jobs of real people.
“If we continue to see an increase in output along the lines of ‘I am not able to answer that question because I am a chat bot,’ we will grow more concerned that regulation is removing power from ChatGPT,” Sarah Hindlian-Bowler, an analyst with investment bank Macquarie, said in a note to clients.
