OpenAI Product Chief Says ChatGPT Will Become Agentic in 2025

The artificial intelligence (AI) industry in 2025 will be defined by the shift to AI agents that move beyond answering questions to completing tasks in the real world. That includes OpenAI‘s plans for ChatGPT.

“Take ChatGPT. This is the year it goes from answering questions for you to doing things for you in the real world,” OpenAI Chief Product Officer Kevin Weil said Tuesday (March 11) during a fireside chat at the Human[X] conference in Las Vegas.

“You can just have it go off and do a bunch of work,” Weil added. “Then it’s not just how efficient can you make yourself, it’s also how many agents can you have (going) off solving problems for you.”

Asked if this meant people would have two-hour workdays since AI agents will be doing tasks, Weil said he believes people will always find things to do.

“I’m not of the opinion that we’re all going to retire to a life of poetry and UBI (universal basic income) and stuff like that. I think people (will) strive,” he said.

“It’s a very human thing to want to solve problems that are bigger than you and strive and try to do hard things,” he added. “People are still going to be people, and we’re going to try and create awesome stuff for the rest of the world to use. We’re just going to have superpowers as we do it.”

What is different about AI is the pace and scope of new innovations that would require a reorientation of one’s strategy.

In the past, “the technology that you’re building with is kind of constant — I mean, a database is a database. It gets a little better every year, but fundamentally, you know what you’re working with. Whereas in AI, it’s just not true,” Weil said.

Now, every two months, “there’s some new model that can do something that computers have never been able to do before, and suddenly you need to rethink your roadmap,” Weil said. “The problems are the same. You can still solve really important problems, but the way you do it, it may be totally different from month to month.”

Read more: Report: Microsoft Considering AI Models to Replace OpenAI’s in Copilot

Weil, who joined OpenAI nine months ago after stints at Twitter (now X), Instagram and satellite imaging company Planet, described his current role as “the most incredible, intense, inspiring, fun, very little sleep” experience of his career. His remarks come as OpenAI continues to dominate the AI landscape with its GPT models, despite increasing competition.

For now, OpenAI is maintaining its lead as the maker of ChatGPT. It currently serves over 400 million weekly ChatGPT users, three million developers using its API, and two million business users on enterprise products, according to Weil.

Weil also noted that the company is iterating models approximately every three months, significantly faster than the previous six- to nine-month cycle, thanks to innovations in reinforcement learning that teach models to reason.

Asked if he thought large language models (LLMs) are getting commoditized, Weil said he didn’t think so.

“People talk about models being commoditized. My personal belief is that that’s actually not true,” Weil said.

It “doesn’t mean that we’re going to have a lead forever. I think those days of us having a 12-month lead are probably gone — there’s just too many smart people, too much going on in the ecosystem,” he said. “But I do think we have a three-to-six month lead, and having a three-to-six month lead is really valuable. We intend to do everything we can to maintain that.”

Weil’s remarks come as Microsoft reportedly is testing other foundation AI models that could replace OpenAI’s models in Copilot. It is also training models that could compete directly with OpenAI, both for use in Copilot and as a base model for developers to build apps on top of them.

Digits CEO Jeff Seibert, who shared the stage, disagreed with Weil that LLMs are not getting commoditized.

“I think you probably will maintain a lead, but the delta will shrink. And for a lot of use cases, it will become relatively commoditized,” Seibert said. “For advanced use cases, yes, there’ll be advantages, but for a lot of (other) stuff, it won’t matter as much” which AI model people choose to use.

Seibert said he advises startup founders to architect their technology such that they can “swap out models very easily, because you want to be able to keep on top of the latest one. And if a new model comes out next week or tomorrow, swap that into your app.”

As for creative professionals worried about AI replacing their jobs, Weil offered reassurance through an anecdote about a film director using Sora, OpenAI’s video generation model.

Rather than replacing the artists, the director found that AI allowed him to rapidly prototype 30 different cut scenes in half a day instead of paying $100,000 for just two options from a special effects studio. The director still used the services of the studio but incurred a lower cost and was able to showcase what he really wanted.

“What results is a much better movie, because he’s had a brainstorm partner that’s infinitely patient and thoughtful,” Weil said.

Addressing safety concerns, Weil emphasized OpenAI’s proactive approach. “I’ve never seen a company that invests as much in safety and integrity ahead of a launch than OpenAI. Before we launch anything, we’re sending new models out to red teamers, we’re engaging with policy folks, in addition to a whole host of internal work.”

Weil acknowledged the company would likely “get some things wrong too,” but stressed their commitment to “iterative deployment and co-evolving with society by launching early and often.”

The post OpenAI Product Chief Says ChatGPT Will Become Agentic in 2025 appeared first on PYMNTS.com.