ChatGPT launched six months ago. Its impact — and fallout — is just beginning | The AI Beat

A mere six months ago today, OpenAI released ChatGPT.

Since then, it’s been a dizzying AI ride: The “interactive, conversational model” became the talk of the AI community within days and a global cultural phenomenon within weeks. In a new, unsettling twist, ChatGPT’s massive popularity can also be tied directly to today’s mainstream media headlines like “AI Poses ‘Risk of Extinction,’ Industry Leaders Warn,” as leaders from top AI labs like OpenAI, Google DeepMind and Anthropic warned in a 22-word statement that “mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war.”

But clearly, the impact — as well as the fallout — of ChatGPT is just beginning. I think it’s worth marking the half-birthday of the world’s most well-known AI chatbot, which got the generative AI hype machine going at full speed. It has already inspired a wave of creative and business applications and become a central part of discussions about ethics, privacy, copyright, data security and misinformation.

ChatGPT has generated comments in Senate hearings, inspired fear in Hollywood, freaked out teachers and gotten New York City lawyers in trouble — but also made millions excited about the opportunities to boost productivity, efficiency, creative ideation and knowledge management. Words like “hallucinations” and “prompt engineering” have become part of the public discourse, while job displacement concerns have exploded, and policymakers have sprinted to try and catch up to the sudden wave of powerful AI development.

ChatGPT was a low-key surprise in November 2022

On November 30, 2022, OpenAI’s announcement was a low-key surprise. At the time, the AI community was mostly talking about what was going on at NeurIPS, a top machine learning and computational neuroscience conference, which was in full swing in New Orleans. There were whispers that details about GPT-4 were going to be revealed there, but instead, OpenAI suddenly announced a new model in the GPT family of AI-powered large language models, text-davinci-003, what it called the “GPT-3.5 series,” that reportedly improved on its predecessors by handling more complex instructions and producing higher-quality, longer-form content.

And then, there it was: After the GPT-3.5 announcement, OpenAI launched an early demo of ChatGPT, another part of the GPT-3.5 series whose dialogue format made it possible “to answer followup questions, admit its mistakes, challenge incorrect premises, and reject inappropriate requests.” 

In a blog post, OpenAI CEO Sam Altman wrote that language interfaces “are going to be a big deal, I think. Talk to the computer (voice or text) and get what you want, for increasingly complex definitions of ‘want’!” He cautioned that it is an early demo with “a lot of limitations — it’s very much a research release.” 

But, he added, “This is something that scifi really got right; until we get neural interfaces, language interfaces are probably the next best thing.”

Altman’s comments immediately sent thousands of AI practitioners to their keyboards to try out the ChatGPT demo and immediately put the tech world in full swoon mode. Within days, Aaron Levie, CEO of Box, tweeted that “ChatGPT is one of those rare moments in technology where you see a glimmer of how everything is going to be different going forward.” Y Combinator cofounder Paul Graham tweeted that “clearly something big is happening.” Alberto Romero, author of The Algorithmic Bridge, calls it “by far, the best chatbot in the world.” And even Elon Musk weighed in, tweeting that ChatGPT is “scary good. We are not far from dangerously strong AI.” 

The hidden danger lurking within ChatGPT

It didn’t take long, however, to identify the hidden danger lurking within ChatGPT: That is, it quickly spits out eloquent, confident responses that often sound plausible and true even if they are not. The model was trained, it was noted, to predict the next word for a given input, not whether a fact is correct. 

By the first week of December, Arvind Narayanan, a computer science professor at Princeton, pointed out in a tweet: “People are excited about using ChatGPT for learning. It’s often very good. But the danger is that you can’t tell when it’s wrong unless you already know the answer. I tried some basic information security questions. In most cases the answers sounded plausible but were in fact BS.” 

Even OpenAI’s Sam Altman admitted ChatGPT’s risks early on. “ChatGPT is incredibly limited, but good enough at some things to create a misleading impression of greatness,” he tweeted on December 10. “It’s a mistake to be relying on it for anything important right now. It’s a preview of progress; we have lots of work to do on robustness and truthfulness.” 

ChatGPT coverage has never slowed

Those risks, however, did not stop the forward march of ChatGPT and other LLMs. By mid-December, experts were saying ChatGPT was having “an iPhone moment.” By January, a top AI conference banned the use of ChatGPT in paper submissions.

In February, ChatGPT competitors like Anthropic’s Claude grabbed some of the spotlight, while Google and Microsoft launched dueling generative AI debuts with Bing and Bard. March brought a wave of chatbot-powered productivity apps, while pioneering AI researcher Yoshua Bengio called ChatGPT a “wake-up call” just in time for OpenAI to move ahead with GPT-4 in yet another surprise announcement. By April, open-source LLMs were having their own moment — and fierce debate.

But nothing seems to be able to unseat the overwhelming popularity of ChatGPT in the public imagination, even though a recent Pew Research survey found that while a majority of Americans have heard of ChatGPT, few have tried it themselves.

Happy half-birthday, ChatGPT. As a large language model, can you plan a proper celebration?