Lori Beer, global chief information officer at JPMorgan Chase, talks about the latest artificial intelligence with the enthusiasm of a convert. She describes AI chatbots like ChatGPT, with their ability to produce everything from poetry to computer programs, as “transformational” and “a paradigm shift.”
But AI isn’t coming to America’s biggest bank soon. JPMorgan blocked its computers from accessing ChatGPT and instructed its 300,000 employees not to put any bank information into chatbots or other generative AI tools.
For now, Beer said, the risk of sensitive data being leaked is too great, and there are too many questions about how the data is used and the accuracy of AI-generated responses. The bank created a closed private network to allow a few hundred engineers and data scientists to experiment with the technology. They are exploring uses such as automating and improving technical support and software development.
The outlook is more or less the same in much of corporate America. Called generative AI, the software behind ChatGPT is seen as a wave of new and exciting technology. But companies across industries are mostly experimenting with the technology and weighing the economic implications. Its widespread use by many companies may still be years away.
According to predictions, generative AI could greatly increase productivity and add trillions of dollars to the global economy. But the lesson that history teaches us, from steam power to the internet, is that there’s often a prolonged gap that passes between the arrival of a major new technology and its widespread adoption, which is what transforms industries and helps feed the economy.
Take the case of the internet. In the 1990s, predictions were made that the internet and the web would revolutionize the retail, advertising and media industries. Those predictions came true, but that happened more than a decade later, well after the dot.com bubble had burst.
During that time, technology has improved and costs have come down, so the bottlenecks have melted away. Over time, broadband internet connections became commonplace. User-friendly payment systems have been developed. Audio and video streaming technology has improved a lot.
The development was fueled by a flood of money and a wave of trial and error by entrepreneurs.
“We’re going to see a similar gold rush this time around,” predicts Vijay Sankaran, chief technology officer at Johnson Controls, a major supplier of construction equipment, software and services. “A lot will be learned.”
The investment frenzy has already begun. In the first half of this year, investments in generative AI startups reached $15.3 billion, almost three times last year’s total, according to PitchBook, which tracks investments in startups.
Technology managers at large companies are testing generative AI software from a range of vendors and watching to see how things play out in the industry.
In November, when ChatGPT was made available to the public, it was a “Netscape moment” for generative AI, said Rob Thomas, IBM’s chief commercial officer, referring to Netscape’s introduction of the browser in 1994. “It breathed life into the internet.” But it was just a start, something that opened the door to new business opportunities that had taken years to explore.
In a recent report, the McKinsey Global Institute, the research arm of the McKinsey consultancy, included a timeline for the widespread adoption of generative AI applications. It started from the premise that currently known technology will be constantly improved, but it did not include a forecast of future major new discoveries. His prediction for widespread adoption was neither short nor precise – it ranged from eight to 27 years.
This wide range is explained by the inclusion of different assumptions about economic cycles, government regulation, corporate cultures and executive decisions.
“We’re not modeling based on the laws of physics — we’re modeling economies and societies, people and companies,” said Michael Chui of the McKinsey Global Institute. “What happens is largely the result of human choices.”
Technology spreads throughout the economy through people, who take their know-how to new industries. A few months ago, Davis Liang left an AI group at Meta to work at Abridge, a healthcare startup that records and summarizes patient appointments for use by its doctors. Its generative AI software enables doctors to save hours of time they would spend typing appointment notes and writing invoices.
Liang, a 29-year-old computer scientist, has authored scientific papers and helped build so-called grand language models that underlie generative AI.
There is a huge demand for your skills these days. Liang declined to reveal how much he earns, but people with his experience and background working at generative AI startups typically earn an annual base salary of more than $200,000, and company shares can push their total compensation much higher.
Liang said Abridge’s key advantage is applying the “super-powerful tool” of AI to the medical industry “to improve the working lives of doctors.” He was recruited by Zachary Lipton, a former research scientist in Amazon’s AI group who is an adjunct professor at Carnegie Mellon University. Lipton joined Abridge this year as Chief Scientific Officer.
“We’re not doing advertising or anything like that,” Lipton said. “It’s hugely fulfilling when you get thank you letters from doctors every day.”
Important new technologies represent sustainable competitive advantages that promote further innovation, leading to the creation of startups that create applications that make the underlying technology useful and accessible. In its early days, the personal computer was seen as a toy for amateurs. But the creation of the spreadsheet program, the cutting-edge application of its day, made the PC an essential business tool.
Sarah Nagy was leading a data science team at Citadel, a giant investment firm, in 2020 when she first tinkered with GPT-3. This was more than two years before OpenAI released ChatGPT. But the power of fundamental technology was already apparent in 2020.
Nagy was especially impressed by the software’s ability to generate computer code from text commands. She thought this would help democratize data analysis in companies, making it broadly accessible to a variety of professionals, not just an elite group.
In 2021 she founded Seek AI to pursue this goal. Today, the New York startup has two dozen clients in the technology, retail and financial industries, working mainly on pilot projects.
Using Seek AI’s software, a retail manager, for example, can enter questions about sales, advertising campaigns and online versus in-store performance to guide marketing strategy and spending. The software then transforms the words into a coded question, searches the company’s data warehouse and presents text answers or retrieves the relevant data.
Nagy said business owners can get responses almost instantly or within a day, rather than two or three weeks as would be the case if they had to request something that needed the attention of a member of a data science team.
“Ultimately, we’re trying to reduce the time it takes to receive a response or access useful data,” Nagy said.
Saving time and streamlining work in companies are the biggest initial goals that most companies pursue with generative AI. New products and services will come later.
This year JPMorgan registered IndexGPT as a possible name for an investment advisory product powered by generative AI.
“This is something we will research and continue to evaluate over time,” said Beer, the bank’s tec director. “But it’s not close to being released yet.”
Translation by Clara Allain