It’s early days in the rise of generative AI such as OpenAI’s ChatGPT and many in the market remain unconvinced of how it will play out for the economy and society, if amazed at its tricks.
Warren Buffett said in a recent interview with Becky Quick on CNBC’s “Squawk Box” that while ChatGPT did “wonderful things” writing a song for him in Spanish, and that “it’s an incredible technological advance in terms of showing what we can do,” he wasn’t convinced about the ultimate outcomes for the world. “I think this is extraordinary but I don’t know if it’s beneficial,” he said.
He did say the time-saving component of the tech is among the things that struck him.
“It can tell you that it’s read every book, every legal opinion. I mean, the amount of time it could save you, if you were doing all kinds of things, is unbelievable,” Buffett said.
That’s where CEOs in the generative AI space are focused.
OpenAI CEO Sam Altman told CNBC’s Julia Boorstin in an interview after being named the No. 1 company on the 2023 CNBC Disruptor list this year that the legal profession is a good example.
“What we’re hearing from customers using our API for legal companies is that it is totally transforming the way they work and the efficiency that any one lawyer can achieve and the accuracy, freeing people up to do more of what they do really well, and having this new tool to sort of give them as much leverage as possible,” Altman said.
That backs up what tech executives working directly with legal firms have previously told CNBC, with one saying of his legal and accounting firm clients that the sentiment right now is not that AI replaces lawyers, but “lawyers using AI are gonna replace lawyers. … Those professionals are going to be more effective, more efficient, they’ll be able to do more,” he said.
“That is a pattern we’re seeing again and again in many industries, and I’m super excited about it,” Altman said. “I do think it will touch almost everything.”
There isn’t much research yet to support these contentions, but early data does support the anecdotal evidence. A study released by MIT researchers in March showed that workers were 37% more efficient using ChatGPT.
Aidan Gomez, CEO of generative AI startup Cohere, which ranked No. 44 on this year’s Disruptor 50 list, pointed to that MIT study in a CNBC interview on Tuesday, saying, “The results are amazing,” he said. “That’s Industrial Revolution-level large. What the steam engine did for mechanical work, mechanical labor, this technology is going to do for intellectual labor.”
Gomez stressed in his comments to CNBC that the research had not yet been peer-reviewed. The authors of the MIT research, Whitney Zhang and Shakked Noy, were unable to comment due to the research currently being in the process of submission to a journal for peer review and publication.
Separately, a new study from researchers at Stanford University and the National Bureau of Economic Research found a 14% productivity increase, on average, when customer service agents for a retail company used generative AI-based conversational assistants.
Generative AI already begun to ‘noticeably impact workers’
Cohere’s platform lets developers and businesses of all sizes — even those without expertise in machine learning — integrate AI features like copywriting, search, conversational AI, summarization or content moderation in their company’s mobile app or service platform. Cohere works with AI customer service tech vendor LivePerson and has cloud deals with Google, Amazon Web Services and Oracle. Salesforce is an investor in the company, one of the first investments the customer relationship management tech giant made this year in a new AI fund. Gomez, along with co-founder Nick Frosst, came from Google Brain, an exploratory deep learning artificial intelligence team that’s now part of Google Research. While at Alphabet‘s Google, Gomez and other researchers helped to develop a new method of natural language processing — transformers — that enable systems to grasp a word’s context more accurately.
Comments like Gomez’s have contributed to the debate about whether AI replaces human labor or augments it. In sectors such as education, those fears are already running high. Gomez, in keeping with the outlook from most AI executives, is sticking to the “augmentative” script.
“What you’re going to see is humans are going to become ten times more effective at what they do,” he said.
He did say we should be wary of companies pointing to AI as the reason for layoffs in the future. He expects that excuse to be made.
But workers also have an advantage, for now, Gomez said: the time it will take to integrate AI technology into the existing technology stack.
“The reality is this will be a slow process over the next half-decade and there will be time to adjust, and change your own job,” he said. “And frankly, you’re going to love it.”
His comments made clear that workers better get used to it.
“We’re pre the real deployment, so I think simmering underneath the water is all this work going on to just transform every product, every single company.”
The MIT study provided more of a mixed assessment of the eventual outcomes for workers and the labor market. The increases in productivity among college-educated professionals performing mid-level professional writing tasks were qualified as “substantial,” and the study showed these workers executed tasks “significantly faster.” Initially low-performing workers, meanwhile, saw output increase and time on task decrease. But the MIT researchers weren’t sure that meant the outlook was good for preserving jobs.
“The experimental evidence suggests that ChatGPT largely substitutes for worker effort rather than complementing workers’ skills, potentially causing a decrease in demand for workers, with adverse distributional effects as capital owners gain at the expense of workers,” they wrote.
The researchers also pointed to limitations in their study. For one, the tasks were “relatively short, self-contained, and lack a dimension of context-specific knowledge, which may inflate our estimates of ChatGPT’s usefulness.” They could not draw conclusions about overall job satisfaction from the results, and, in capturing “only direct, immediate effects of ChatGPT on the selected occupations” they cannot account for many other factors that will weigh in labor markets and production systems as they adapt to new technologies like ChatGPT, or how AI will influence each occupation, task, and skill level.
The only conclusion they made with confidence in their paper: “For now, the evidence we provide suggests that generative AI technologies will — and have already begun — to noticeably impact workers.”
Watch the full CNBC Disruptor 50 interview above for more of this leading generative AI CEO’s views on how the next few years of work will play out.