April 8th, Despite the rapid advancement of generative artificial intelligence such as ChatGPT, many enterprises remain cautious about adopting this technology, opting for a wait-and-see approach.
Take Reynolds American, the U.S. tobacco company, for example. It is taking slow steps, dedicating limited resources to test artificial intelligence. Aaron Gwinner, Chief Information Officer of Reynolds, stated that the company is seeking to leverage AI to enhance the analysis of large datasets but is not deploying the technology in high-risk financial applications or customer-facing roles.
"Inherent risks come with generative AI," said Gwinner. "Before fully implementing AI projects, it's crucial to establish principles and build a solid foundation."
Microsoft's Copilot, Google's Gemini, and OpenAI's ChatGPT are all striving to introduce generative AI technology into the business domain, with supporters hoping for revolutionary impacts. For instance, in February of this year, the Swedish fintech company Klarna stated that ChatGPT tools can now handle the workload equivalent to 700 customer service representatives.
However, many companies have been slow to adopt this technology.
Surveys indicate that many enterprises, like Reynolds, are proceeding cautiously. According to data released by the U.S. Census Bureau in March of this year, only about 5.4% of businesses are utilizing various forms of artificial intelligence to produce goods or provide services. Furthermore, statistics suggest that the information industry and large enterprises are adopting AI at a higher rate than other sectors.
According to a report released by market research firm Gartner in March of this year, the industries most likely to claim they will adopt generative AI are life science companies, oil and gas, and communication service providers, with 42% to 49% of companies expressing such plans. However, many surveys also indicate that despite most companies expressing intentions to adopt AI, they have not taken actual steps, a phenomenon Gartner refers to as the "ambition-achievement gap."
Daniel Colson, co-founder of the AI Policy Institute, stated that organizations are still evaluating the practical applications of generative AI.
"The issue has long been unclear. I believe this has led to uncertainty about the true value of large language models in commercial applications," Colson commented on the AI systems behind ChatGPT and other tools.
Some are concerned that the investment in generative AI does not yield proportional returns. Many organizations struggle to find suitable use cases.
Luminos Law, a law firm specializing in AI risk management consulting for clients, has considered using generative AI itself but has made no progress to date.
Andrew Burt, co-founder of Luminos Law, stated, "We've discussed it, but the fact is the real value of generative AI lies in providing knowledge and insights on a large scale to the public." "Our scale is small, and our services are quite specialized, so this path isn't suitable for us."
Last year, online mental health service company Koko revealed it had started using ChatGPT to compose empathetic messages to users. As an early adopter of ChatGPT, Koko's approach sparked some controversy.
According to Rob Morris, CEO of Koko, despite the messages being ultimately reviewed by humans and sent faster than expected, users still expressed dissatisfaction upon learning they came from a virtual companion.
"Virtual sympathy feels both strange and hollow," Morris said. He also stated that the company's use of ChatGPT was approved by an external board but discontinued after better use cases were not found.
Mind Meld PR, a public relations agency based in Vancouver, sends press releases to journalists as part of its work. The company stated that the process of trying to use generative AI did not go as smoothly as expected because the content required constant refinement, consuming time comparable to fully manual operations.
Jonathan Narvey, CEO of Mind Meld PR, stated that AI sometimes provides very precise answers to research questions, but they later realized the results generated by the tool were not authentic.
Narvey admitted, "We tried, and it was a complete disaster."
During the ChatGPT craze, the phenomenon of generative AI providing precise but fictitious answers is widespread.
In February of this year, a Canadian airline was forced to refund a customer due to incorrect information provided by a chatbot. The airline stated that the technology behind the chatbot was more outdated and simplistic than current AI technology, but this incident highlighted the risks of providing information to customers through software. The Canadian airline subsequently withdrew the chatbot from service.
Although large enterprises are more likely than small businesses to adopt generative AI, they also recognize the potential security risks. A recent survey conducted by Cisco among privacy and security professionals showed that 92% of respondents believe generative AI is fundamentally different from other technologies, requiring new technologies to manage data and risks. More than a quarter of respondents even prohibit the use of this technology in their work. Time could be the primary factor limiting the widespread adoption of this technology. Search engines and social media, as breakthroughs of the internet age, took years to become ubiquitous. According to data from the Pew Research Center, in 2002, just four years after Google's founding, only 29% of internet users used search engines daily. In contrast, ChatGPT, launched in November 2022, amassed 100 million monthly active users in just two months, becoming the fastest-growing application in history at that time, as per analysts at UBS.
Generative AI may eventually follow a similar trajectory to early technologies. Microsoft's research found that 77% of users who tried the generative AI product Copilot expressed a reluctance to give it up.
Reynolds of tobacco company Reynolds says the company is trying to strike a balance between the competitive advantage that rapid action might bring and the risks of moving too quickly.
"If you look back over the past century of technological revolutions, this will be one of the biggest," Reynolds said. "But we need time to prepare."
