For several weeks, the 38 employees at advertising agency Struck had the same conversation—over coffee, on Slack, inside the kitchen of its downtown Salt Lake City office: “What the heck does ChatGPT mean for our industry, our company, and our jobs?”
“There are a few here and there who are nervous, but the general consensus is that it’s not going to totally replace us,” says Tim Behuniak, a strategist at the 25-year-old agency. “We can use it like a tool in our library, just like Photoshop.”
It’s been a common conversation since startup OpenAI revealed its cutting-edge language generation model ChatGPT in January and immediately drew 1 million users. Smarter than any other chatbot that came before it, the artificial intelligence bot can answer virtually any question, explaining quantum mechanics and writing a poem in the style of Shakespeare about…your cat, for instance.
“It blew people away when it came out,” says Dan Case, who follows AI trends as a partner with Cicero Group, a management consulting firm in Salt Lake City. “People were hit with the realization that they are entering a new era of technology, and one where artificial intelligence would be real, not just an idea.”
ChatGPT is showing tremendous potential for revolutionizing the way we interact with technology. The artificial intelligence model, which is trained on a massive dataset of text from the internet, is able to generate human-like responses to a wide range of prompts. This makes it ideal for applications such as chatbots, virtual assistants and language translation. ChatGPT has been able to successfully carry out conversations on a variety of topics, including news, sports and current events. It can also understand context and maintain a consistent tone and persona in its responses, making it highly engaging and user-friendly.
That last paragraph? ChatGPT wrote it. The bot wrote far more than one paragraph—an entire article, in fact—and it took three seconds to spit it out after receiving my request to write a news article about ChatGPT. We’ll write our own story; thank you.
The first peek at the transformational power of AI was unveiled in 2014 when Google released DeepDream, which identified patterns in art to create psychedelic images. OpenAI’s 2021 release of DALL-E then captivated the public’s imagination with its impressive text-to-image AI model that creates original digital artwork out of imaginative text prompts. Since then, a number of other companies, such as Google, Microsoft and Stable Diffusion, have released their own versions.
ChatGPT’s vast knowledge is thanks to the huge flood of content that’s been amassed on the internet over the last decade, all of which is ripe for data-crunching AI training models, says Case of Cicero Group. ChatGPT uses a large language model algorithm that relies on a deep learning neural network to process the natural language. These algorithms are essentially fine-tuned to predict the next word in a sentence while relying on large swaths of data to understand and maintain relevant context in its prediction. As the algorithms get more sophisticated, they will only get smarter. “The next 10 years are going to be an AI renaissance,” Case says.
The tech behind ChatGPT is expected to shift the landscape for the written word dramatically. It will enable companies to use AI to debug software, deliver sales pitches, draft skeletons for white papers, stuff content with SEO keywords, write blog posts, create training manuals and more. The bot’s capabilities are pretty mind-blowing: it even passed a Wharton MBA exam.Education and homework will also look a lot different going forward. Students are already asking ChatGPT to write their essays and do their homework. Obligingly, GPT will spit out a paper within seconds on any topic, from Genghis Khan to the main themes of Neoclassicism. Because of this, we may be moving to a time of oral exams, proctored exams, and new kinds of verification and encryption. “As it improves, homework and papers will be a thing of the past,” said Nicholas Thompson, CEO of the Atlantic, in a LinkedIn post.
In that same vein, job candidates could cheat on hiring tests using ChatGPT to answer questions or, in the case of a computer programming candidate, to write code for automated tests, says Lindsey Zuloaga, chief data scientist at Salt Lake City-based HireVue, which uses AI to analyze video interviews for employers. Even before ChatGPT, there was an arms race for algorithms to detect cheating in online hiring exams. Hirevue’s CodeVue software, for instance, can flag a candidate who pauses for long periods of time and note when text or lines of code look too similar to others. Expect algorithms to pop up to identify text created by ChatGPT, she says, since the bot tends to write in a particular style.
ChatGPT will also change the way we search the internet, Thompson said in another post. People may use OpenAI’s ChatGPT to search for answers to questions such as, “What is a low-sodium diet?” Rather than being steered to any number of websites, they would receive an answer drawn from a conglomeration of information from various websites.
“A lot of Google search traffic will change,” Thompson explained. “There are a lot of businesses that are going to have to evolve very quickly because their principal source of traffic is going to change or go away or shift. It’s not going to happen immediately. But it’s going to happen.”
This technology is also poised to fundamentally change how we do business. Employees are already sitting down to work each day, opening up ChatGPT in the morning, and using it all day for research, summarizing content or preparing job interview questions. Creatives are asking it to write screenplays and poetry, and social media strategists are using it to crank out content—all within seconds and by scraping the vast data set called the internet, indexed up to 2021. If you try to log onto ChatGPT now, there’s a good chance the chatbot will tell you that it’s overwhelmed by demand and to come back later.
As many as 8 percent of open jobs tallying up to 800,000—including editors, reporters, customer service agents and quality assurance engineers—will immediately be impacted, according to data by the Josh Bersin Company, which analyzes talent market trends. While these jobs may not immediately disappear, ChatGPT will change how they’re done, says Josh Bersin, the firm’s CEO. For now, people expect the tech to be a time-saver: that mindless intern who can do market research, come up with a boilerplate for a proposal or write a pretty vanilla story.
There are plenty of problems with it. ChatGPT synthesizes things that people have already written. Sometimes it’s smart. Sometimes it’s not. It doesn’t always know the connections between words or ideas.
“It’s a highly intelligent cut and copy machine,” Bersin says. “It looks intelligent, but under the covers, there’s a lot of things it doesn’t know. It’s only as smart as the data sources it has indexed.”ChatGPT users don’t really know where the information came from and can’t necessarily ensure it’s authoritative, accurate and complete. There’s also an inherent potential for it to be biased if the data it accesses has bias. There’s no fact-checking or attribution, and the information could be copyrighted or plagiarized. It is very good at justifying its answers. “It could be flat-out wrong,” Case says.
An even bigger concern? This kind of AI might pose an imminent threat to society. Gary Marcus, emeritus professor of psychology and neural science at New York University, warns that AI systems like these may wind up being misinformation war machines. Already, some chatbots—not ChatGPT—will write misinformation for you, even citing non-existent studies as evidence. Eventually, he says, we won’t be able to tell the difference between truth and lies. “We have no existing technology that really protects us from the onslaught, the incredible tidal wave, of potential misinformation like this,” he says. He calls it AI’s “Jurassic Park” moment.
Yet, companies are aggressively investing in the technology. In January, Microsoft, which first backed OpenAI in 2019, said it would make a multibillion-dollar investment in OpenAI. At least six companies—with names like Adept AI, AI 21 Labs, Anthropic, Character, Cohere and Inflection—are building competing AI chatbots. Google Bard, Google’s own experimental conversational AI to rival ChatGPT, was announced in early February.
Company leaders everywhere are chewing on how they, too, can use this advanced AI for their own purposes. Bersin compares ChatGPT to what happened when the first iPhone was released, and everyone wanted to build apps but wasn’t sure how. “Companies are going to go nuts [with the technology],” Bersin says. “They’re going to play with it and see what works.”
OpenAI has an API for ChatGPT that would allow companies to leverage that same AI to perform tasks and answer customer queries. But for a company to rely on ChatGPT in a significant way, it must source valid, deep and expansive domain data, Bersin says. The best AI chatbots will provide internet users with specific answers drawn from internal expertise on narrow verticals and topics. They will understand the user and the domain of the conversation deeply. For example, a great recruiting chatbot might never be good at leadership coaching, and vice versa.
Company leaders are also building internal chatbots that could allow employees to quickly get answers to questions, whether it’s finding a certain company document or looking to find an employee with specific expertise. It could be a huge time-saver: Employees spend at least two hours a day just looking for documents, information or people they need to get their jobs done, according to Glean’s “Hybrid Workplace Habits & Hangups” report. The report states that nearly half of employees get so frustrated searching for internal information that they’d consider leaving a job over it.
ChatGPT’s capabilities have been a hot topic at the 400-person HireVue in Utah. In hiring, there’s potential for ChatGPT to summarize a person’s LinkedIn profile or resume and even make assumptions about their skill sets—inferring a person probably knows physics even if it’s not explicitly stated on their resume, for instance. It might also help develop interview questions for candidates outside a leader’s normal expertise, such as an attorney or a computer programmer. “We spent a lot of time in the last month exploring it,” Zuloaga says. “It’s prone to make mistakes, so a lot more testing is needed.”
In Park City, Bassam Salem has been playing with how the technology can be used in customer service chatbots. Late last year, New Jersey customer experience conglomerate NICE acquired his firm, AtlasRTX, which builds automated chatbots, mainly for clients in home building and higher education. The bots can converse with their web visitors about everything from home prices to the benefits of various majors, but the conversations are limited in scope.
The one thing NICE bots can’t do is converse in a human way that extends beyond information about the specific industry in which they were trained. Eventually, NICE could use the ChatGPT API in its own chatbots and enable what Bassam calls “chit-chat capabilities.” Because ChatGPT is trained on data up to 2021, it doesn’t produce real-time answers, such as the current weather in a particular city, for example. But one day, it may. “It could make those conversations a lot more intriguing,” Salem says.
Meanwhile, 30 miles away, Struck managers have been holding meetings inside the ad agency to figure out how to leverage GPT when it comes to mundane tasks, such as using it to search for Google images for a mood board or to aggregate market data faster.
Behuniak believes that ChatGPT won’t eliminate the creative process. Those novel ideas behind that ultimate tagline or core positioning strategy, he says, tend to come when a person’s mind wanders while on a walk or in those ah-ha moments of a conversation with a colleague. “There are certain subtleties and nuances of being human that are hard for AI to capture,” Behuniak says.
For now.