Advancements in artificial intelligence (AI) are happening so fast that it’s hard to keep track —and even harder to predict what’s next.
But Oracle Cloud Infrastructure’s SVP of AI and Data Management Greg Pavlik cut through the noise, flagging two use cases that could have the most impact on the future: chain of thought and iterative reasoning.
A chain of thought is a "sequence of connected ideas or reasoning that forms a mental process. It's how one thought leads to another in a logical progression," Sid Nag, VP of Gartner's Technology and Service Provider Group, explained to us via email.
"Iterative reasoning on the other hand involves a process of repetitive refinement or improvement through successive iterations. It's a method where a solution or understanding is developed gradually, often by refining and revisiting ideas or processes," Nag said.
Both of these use cases are complex developments within the world of language models, according to Pavlik. But in addition to pulling more relevant information for users than today's systems can, they will also be able to "act on behalf of the user leveraging that information,” he said.
Today’s AI models, generally just spit out answers based on prompts users submit. But Pavlik said a chain of thought system with iterative reasoning would include an orchestration layer that uses AI to decompose the prompt into a series of actions or steps.
So, for instance, a user could tell the system “I want to take a flight to Los Angeles for a meeting. What flights are available? Which are the safest? And can you help me book one?” The system would then work to satisfy each of those requests.
Basically, Pavlik said, in the future such AI systems could act like knowledge workers, not just getting information for users, but also acting on their behalf and gradually growing smarter. Think Google Duplex, but way, way more advanced and versatile.
Just how far in the future is this development? It’s hard to say given how quickly the AI landscape is changing.
“The technology seems to make some sort of major breakthrough one to two times a year where there’s a major inflection point. But it’s very difficult to say with certainty are we a year away, are we three years away, are we five years away. The reality is we don’t know,” Pavlik said.
Go with the AI flow
Pavlik, who has been at Oracle for the better part of two decades, said that kind of uncertainty is something he has had to get used to over the course of his career.
“As much as I think everyone wants to have stability in their job and be in a kind of a comfort zone, the tech industry is just not like that,” he said. “It’s almost like a series of creative destruction waves. You just need to be prepared for that.”
The exec said while he knew this intellectually, it was quite jarring to experience firsthand. Back in 2012, Pavlik left Oracle to work at big data company Hortonworks. The company made it through an IPO and had a solid plan to scale. But then cloud storage exploded onto the scene “and everything pivoted.” Its outlook changed so much, in fact, that Hortonworks merged with Cloudera in 2018.
“It was a real awakening in terms of how quickly tech can change and how disruptive those changes can be. We were riding high on this company trajectory and then within a year everything turned over,” he said.
“It was a wake-up call that when you’re in tech you need to be nimble and you need to be ready for change," said Pavlik.
That idea – be prepared – carries over into Pavlik’s personal life. The exec is a brown belt in Brazilian Jiu Jitsu and makes a point of heading “into the mountains” at least once a year. There, the consequences of being unprepared can be extreme.
Last year during a trip to Alaska, for instance, Pavlik said he left camp without his rifle to get water. On the way back he crossed paths with a massive bear.
“It immediately saw me and started lumbering towards me,” he recalled. “I ran up to the top of the mountain and tried to use the GPS to figure out how to get back to camp. But that was a lesson, just make sure you have bear spray or some sort of protection.”
That lesson — be prepared — is a common thread in Pavlik's life. At work, he and the Oracle team are working to prepare for an influx of demand for AI technology that is expected to come from enterprises.
Last week, for instance, the company launched a fully managed OCI Generative AI service which lets enterprises tap models from Cohere and Meta for text generation, summarization and semantic similarity tasks.
Deep roots
On the tech front, Pavlik’s lessons began early. He told Silverlinings his dad was an electrical and computer engineer in the military and they built computers at home together. He grew up reading IEEE Spectrum (!). He also took programming courses in middle school and moved through the rest of his education with the intention of working in the sciences.
Working at Oracle has been an interesting challenge, he said. Being a late mover in the cloud arena gave Oracle a chance to rethink the decisions of early players and correct the decisions they thought those companies got wrong, he noted.
“We’ve applied that at every layer of the stack. Networking, compute, all the way up through the way we’ve thought about things like big data,” he said. “It’s been really interesting and kind of a unique opportunity to think differently about the cloud and, we think, do it better.”
As far as AI is concerned, Pavlik said he thinks it’ll have as big an impact on the world as the internet or cloud did. He noted retail is “fundamentally different” than it was 20 years ago thanks to the internet. In the same way, he thinks AI’s impact will extend well beyond the tech arena.
Read more inspiring profiles of tech leaders here.