ChatGPT, OpenAI’s viral AI chatbot, turns one today.
A year ago, OpenAI released ChatGPT as a “short key research preview” — reportedly inspired in part by a bitter feud with AI startup Anthropic. The goal, the leadership of OpenAI SPOKE the OpenAI rank-and-file at the time, was to gather more data on how people use and interact with generative AI to inform the development of future OpenAI models.
Initially a basic free-to-use, web-based and chat-focused interface on top of one of OpenAI’s existing models, GPT-3.5, ChatGPT will go on to become the company’s most popular product… fastest growing consumer app in history.
In the months following its launch, ChatGPT gained paid tiers with additional features, including a plan aimed at business customers. OpenAI also upgraded ChatGPT with web search, document analysis and image generation (via DALL-E 3) capabilities. And, relying on speech recognition, voice synthesis and text-image understanding models developed in-house, OpenAI gives ChatGPT the ability to “hear,” “speak,” “see” and act.
In fact, ChatGPT has become OpenAI’s number one priority – not just a product but a development platform to build on. And, as often happens in a competition-driven market, it shifted the focus to other AI companies and research labs, too.
Google scrambled to launch an answer to ChatGPT, eventually releasing Bard, a more or less identical AI chatbot, in February. Countless other ChatGPT rivals and derivatives have hit the market since, the latest being Amazon Q, a more business-oriented take on ChatGPT. DeepMind, Google’s main AI research lab, is expected to debut a next-generation chatbot, Gemini, before the end of the year.
Stella Biderman, an AI researcher at Booz Allen Hamilton and the open research group EleutherAI, told me that she doesn’t see ChatGPT as an AI breakthrough per se. (OpenAI, which has released several research papers about its models, which ChatGPT never released). But, he says, ChatGPT is a bona fide “user experience improvement” – that takes AI at its core.
“The main effect (ChatGPT) has (is) to encourage people who train AIs to try to imitate it, or to encourage people who study AI to use it as their central object of study,” Biderman said. “You used to have to be skilled, even if not an expert, to always get useful things out of (text-generating models). Now that it’s changed … (ChatGPT) brings a huge amount of attention and discussion about technology.
And ChatGPT is still getting a lot of attention – at least if third-party statistics are anything to go by.
According to Similarweb, the web metrics company, OpenAI’s ChatGPT web portal saw 140.7 million unique visitors in October while the ChatGPT iOS and Android apps had 4.9 million monthly active users in the US alone. Data from analytics firm Data.ai suggests the apps have generated nearly $30 million in subscription revenue — a huge amount considering they launched just a few months ago.
One of the reasons for ChatGTP’s enduring popularity is its ability to conduct conversations that are “convincingly authentic,” according to Ruoxi Shang, a third-year Ph.D. student at the University of Washington studying human-AI interaction. Before ChatGPT, people were already familiar with chatbots — they had been around for decades. But the models using ChatGPT are much more sophisticated than what many users are used to.
“Human-computer interaction researchers have learned how conversational interfaces can improve the understanding of information, and the socialization aspects of chatbots bring more engagement,” Shang said. “Today, AI models enable conversational agents to conduct conversations that are almost indistinguishable from human dialogues.”
Adam Hyland, also a Ph.D. student studying AI at the University of Washington, focuses on the emotional side: ChatGPT conversations have a different “feel” than other non-standard chatbots.
“In the 1960s, ELIZA offered a chatbot, the answer was very similar to how people reacted to ChatGPT,” Hyland said, referring to the chatbot created by MIT computer scientist Joseph Weizenbaum in 1966. “People who interact with the system are known to be emotional. enter a narrative through a line of chat messages.”
Indeed, ChatGPT has impressed cynics like the New York Times’ Kevin Roose, who called it’s “the best AI chatbot ever released to the public.” In The Atlantic magazine’s “Breakthroughs of the Year” for 2022, Derek Thompson included ChatGPT as part of the “generative-AI eruption” that “may change our thinking about how we work, how we think and what human creativity is.”
ChatGPT’s skills go beyond conversation, of course – another likely reason for its staying power. ChatGPT can complete and debug code, create music and essays, answer test questions, generate business ideas, write poetry and song lyrics, translate and summarize text and even yet. follow a computer running Linux.
An MIT study showed that, for tasks such as writing cover letters, “dangerous” emails and cost-benefit analysis, ChatGPT reduced the amount of time workers needed to complete tasks by 40% while increasing output quality by 18%, as measured by a third party. evaluators.
“Because (the AI models that power OpenAI) are trained on a lot of data,” Shang added, “they shift the focus from training specialized chatbots for specific domains to create many general purpose systems that can manage different topics easily by prompting with instructions… (Chatbots like ChatGPT) do not require users to learn anything new form of language, as long as they provide a task and some desired output like how a manager talks to an intern.
Today, there is mixed evidence as to whether ChatGPT is actually used in these ways. A Pew Research survey from August showed that only 18% of Americans have tried ChatGPT, and that most who have tried it use the chatbot for entertainment purposes or to answer a question. Teenagers probably don’t EXERCISES ChatGPT all the time, however (despite what some alarmist headlines means), with a poll finding that only two out of five teenagers have used technology in the past six months.
ChatGPT’s limitations may be to blame.
Although undeniably capable, ChatGPT is far from perfect due to the way it was developed and “taught.” Trained to guess the likely next word — or likeist next parts of words — by observing billions of examples of text from around the web, ChatGPT sometimes “hallucinates,” or writes answers that seem plausible but not really right. (chatGPT’s hallucinating tendencies got its answers banned from the Q&A site Stack Overflow and from at least one academic conference — and accused of blasphemy.) ChatGPT may also show bias in its responses, responding to sexist and racistovertly Anglocentric ways — or regurgitating portions of the data it was trained on.
Lawyers Granted after using ChatGPT to help write the motions, discovering — too late — that ChatGPT had invented fake citations in the case. and scores of authors sued OpenAI over a chatbot that regurgitated parts of their work — and received no compensation for it.
So what’s next? What will ChatGPT’s second year do, if not the same?
Interestingly – and fortunately – some of the more dire predictions about ChatGPT did not come true. Some researchers fear that the chatbot will be used to generate disinformation on a large scale, while others sound the alarm on ChatGTP’s phishing email-, spam- and malware-generating potential.
The concerns endorsed European policymakers to mandate security assessments for any products using generative AI systems like ChatGPT, and more than 20,000 signatories – including Elon Musk and Apple co-founder Steve Wozniak – to sign in an open letter calling for an immediate halt to large-scale AI experiments like ChatGPT.
With the launch of GPTs, OpenAI’s tool for building custom conversational, action-capturing AI systems powered by OpenAI models, including the models that underpin ChatGPT, ChatGPT will become more a gateway to a wider ecosystem of AI-powered chatbots rather than the end-all-be-all.
With GTPs, a user can train a model on a collection of cookbooks, for example, so that it can answer questions about the ingredients for a specific recipe. Or they can provide a model of their company’s proprietary codebases so that developers can check their style or create code according to best practices.
Some of the initial GPTs – all created by OpenAI – include a Gen Z meme translator, a coloring book and sticker creator, a data visualizer, a board game explainer and a creative writing coach. Now, ChatGPT can perform these tasks given careful prompting and foreknowledge. But purpose-built GTPs oversimplify things – and may just kill the cottage industry that has sprung up around creating and editing prompts to feed into ChatGPT.
GPTs introduce a level of personalization far beyond what ChatGPT currently offers, and – once OpenAI resolves its capacity issues – I expect we’ll see an explosion of creativity there. Can ChatGPT be seen as before after the flood of GPTs in the market? Probably not. But it won’t go away — it’ll just adapt and evolve, no doubt in ways that even its creators can’t anticipate.