Released to the public in late November, ChatGPT, the highly articulate AI program, has already created quite a stir, with reactions ranging from the awestruck to the appalled. The New York Times deemed ChatGPT “the best artificial intelligence chatbot ever released to the general public.” Others have zeroed in on concerns over the bot’s potential effects on education and cybersecurity.
Amid this flurry of news, takes, and predictions, it’s become easy to lose sight of the basics of ChatGPT. Find out what you can’t miss about ChatGBPT, its potential influence on search, and its immediate impact on your brand in this quick guide to the disruptive new technology.
What is ChatGPT?
Developed by San Francisco-based research lab OpenAI, ChatGPT is what’s known as a large language model, or LLM, chatbot. Powered by artificial intelligence, ChatGPT is capable of emulating human language and naturally engaging in dialogue in response to users’ prompts.
In the first week after its release, its remarkable array of language skills went viral. And as users across the internet shared examples of ChatGPT’s ability to compose emails, haikus, term papers, and even standup comedy bits in response to their prompts, the platform quickly gained over a million users.
How does ChatGPT Work?
Just how is the application able to ape the style of Hemingway just as well as it churns out computer code? It starts by doing some reading.
LLMs like ChatGPT are trained on huge amounts of language data. In Chat GPT’s case, this pre-training incorporated 570GB of data from the internet—writing ranging from books and articles to Reddit conversations and Wikipedia. All told, the system fed on over 300 billion words. The result of all this training: an incredible ability to understand the patterns and structures of natural language.
By training on this data, LLMs like ChatGPT become better and better at predicting the next word in a sentence—and, eventually, the next sentences, paragraphs, and even pages. In this way, LLMs are like souped-up versions of autocomplete technology like Google’s Smart Compose for Gmail. ChatGPT, however, is no ordinary LLM.
What Makes ChatGPT Different
While many LLMs’ predictive abilities become quite powerful as they’re trained on larger datasets, ChatGPT’s breakthrough is driven by the use of human feedback during training.
Employing a technique called Reinforcement Learning with Human Feedback (RLHF), ChatGPT was trained not only in predicting the next word in a string of text but in delivering output that’s based on the intent and needs of the human input it receives. Thanks to this, ChatGPT is capable of much more frequently delivering answers that seem satisfactory.
Seem is an important word here, though. Because what the current iteration of ChatGPT actually excels at is largely imitation, not actual comprehension or subject-matter mastery. As critics of the software point out, and OpenAI themselves readily admit, with no sense of truth guiding the system’s parameters, “ChatGPT sometimes writes plausible-sounding but incorrect or nonsensical answers.” What’s more, the company says it may produce “harmful instructions or biased information.”
The Future of Search with ChatGPT
Nevertheless, ChatGPT’s ability to respond so articulately and flexibly to user queries has led some to call it the future of search—and it’s not hard to see why.
Rather than a page of links that users need to sift through and evaluate, ChatGPT queries turn up succinct, clearly written sentences. And its versatility is also clear, with the technology capable of generating answers, ideas, and suggestions on everything from formal business plans to Christmas gifts.
Indeed, at Google, the search giant’s leadership declared a “code red” after the release of ChatGPT. Despite significant investments in its own similar AI chatbots, the technology holds the potential to upend Google’s entire business model, which so heavily relies on the ad revenue generated by search activity.
Meanwhile, Microsoft has shown no signs of hesitating. Having already invested $1 billion in OpenAI in 2019, they’re reportedly in talks to invest a further $10 billion.
Plus, in a direct bid to compete with Google Search’s Knowledge Graph, a ChatGPT-powered version of Bing is said to be on the way. While it’s still unknown how exactly ChatGPT will be incorporated into the Microsoft engine, the company hopes to release the feature by the end of March.
What Brands Need to Know Today
As Microsoft’s moves in the market make clear, it won’t be long before ChatGPT—or chatbots just like it—are powering a search engine and affecting how consumers find brands. And though there’s plenty we still don’t know about what exactly that will look like, it will be well worth keeping an eye on.
In the meantime, when its servers aren’t crashing due to the surges in traffic, ChatGPT already offers a range of use cases to support day-to-day tasks. For example, the program can be used to save time while performing market research, writing complex Excel and Google Sheets formulas, and more—and it may not be long until the tool is augmenting apps like Word, PowerPoint, and Outlook.
Precision's Thoughts
Given the conversation-style user input, when ChatGPT is utilized in a search engine capacity, it will be processing long tail "searches". That means content remains king. Brands should ensure that they have as much content related to their subject area, including new trending content, as possible.