Sam Altman, chief executive officer of OpenAI, on the Hope Global Forums annual meeting in Atlanta, Georgia, US, on Monday, Dec. 11, 2023.
Dustin Chambers | Bloomberg | Getty Images
DAVOS, Switzerland — The rise of generative artificial intelligence has dominated private and public discussion on the World Economic Forum as the world’s largest technology companies, including Salesforce, Microsoft and Google, temporarily take over local storefronts as a show of force to conference attendees.
While 2023 highlighted the technology’s ability, 2024 will probably be about increasing the accuracy of results in order that executives in high-stakes industries such as hospitals and manufacturing can get comfortable using AI, said Intel Chief Executive Officer Pat Gelsinger in a CNBC interview.
“You have now reached the top of today’s AI utility,” Gelsinger said. “This next phase of AI, I think, will probably be about constructing formal correctness into the underlying models.”
Whether it’s doctors relying on artificial general intelligence (AGI) for diagnoses, warehouses using it to examine for assembly line malfunctions, or automated driving, humans have to get more comfortable with the accuracy of the technology, Gelsinger said.
“Certain problems are well solved today in AI, but there’s a number of problems that are not,” Gelsinger said. “Basic prediction, detection, visual language, those are solved problems right away. There’s a complete lot of other problems that are not solved. How do you prove that a big language model is definitely right? There’s a number of errors today. So you continue to need you already know, essentially, I’m improving the productivity of a knowledge employee. But at the top of the day, I would like the knowledge employee to say is it right.”
Pat Gelsinger, CEO Intel, speaking on CNBC’s Squawk Box on the WEF Annual Meeting in Davos, Switzerland on Jan. sixteenth, 2024.
Adam Galici | CNBC
The most effective strategy to improve accuracy is thru experimentation and co-piloting tests to advance adoption, said Clara Shih, CEO of Salesforce AI. The AI can adjust to different standard deviation confidence levels as users get comfortable that the technology might be trusted in high-stakes scenarios, Shih said in an interview.
Three phases of AGI will guide adoption, Shih said. Phase one is actively using the technology as an assist for work. Phase two is consciously watching the technology in its autopilot mode to make sure accuracy. The ultimate phase is letting go and trusting the technology will work to the boldness level of 1’s selection, Shih said.
“You may tell the AI to be conservative for higher stakes until a human co-pilot essentially graduates it to autopilot,” Shih explained.
The three-phase approach, which heavily draws upon human acceptance of the reliability of the technology, should make AGI less scary than some have speculated, said Open AI CEO Sam Altman during a panel discussion Tuesday with Bloomberg’s Brad Stone.
“That is far more of a tool than I expected,” Altman said. “It’ll recover, nevertheless it’s not yet replacing jobs. It is that this incredible tool for productivity. It is a tool that magnifies what humans do, lets people do their jobs higher and lets the AI do parts of jobs.”
The past yr has been “a yr of discovery,” said Nasdaq CEO Adena Friedman in an interview. The financial industry, including Nasdaq, will use AI to update old code to modernize aging systems, improving automated workflows, which may save employees hours every day, Friedman said.
“It got here onto the scene somewhat over a yr ago,” Friedman said. “We did some experimentation. We began to sort of understand the potential of it. This yr will probably be the yr of activations for us and for everybody.”
WATCH: Nasdaq CEO Adena Friedman discusses economy, opportunities in AI at Davos