- Brad Lightcap, chief operating officer at OpenAI, said that current generative AI will be “laughably bad” within a year.
- The executive predicts that ChatGPT could soon take on more “complex tasks” and become a “great teammate.”
- The tech giant plans to release its latest language model, GPT-5, as early as this summer.
If you think ChatGPT is great, imagine what AI chatbots will be able to do in a year's time.
One OpenAI executive predicts that generative AI tools like his company's today will be insignificant compared to the future.
“Within the next 12 months, the systems we currently use will become laughable. It’s going to be terrible.” “We think we are moving towards a world where they are more capable.”
Lightcap says the large-scale language models that people use to get work done and achieve personal goals will soon be able to handle “more complex tasks.”
Reached through OpenAI, the COO did not immediately respond to a request for comment from Business Insider before publication, asking what the so-called “complex work” meant.
He added that AI will have a more “systemic relationship” with users, meaning the technology will act as a “great teammate” who can help users with “any problem.” .
“It's going to be a different way to use the software,” an OpenAI executive said during a panel discussion regarding AI's predictive capabilities.
Given his predictions, Lightcap acknowledged that it may be difficult for people to “really understand” and “internalize” what a world with robot assistants would be like.
But in the next decade, COOs believe that talking to AI the same way you would talk to a friend, teammate, or project collaborator will become the new norm.
“I think this is a significant change that we haven't fully grasped yet,” he said of the outlook 10 years from now.
The executive's thoughts seem to hint at what's to come with GPT-5, the latest model of OpenAI that the AI giant plans to release as early as this summer.
One CEO who received a demo of OpenAI's new model told Business Insider on condition of anonymity that GPT-5 is “very good” and “significantly better” than GPT-4, the LLM behind ChatGPT. “There is,” he said.
The CEO didn't provide details, but said he hinted at other “undisclosed features,” including the ability for OpenAI to summon AI agents to perform tasks on its own.
Still, it is clear that AI has a long way to go before it can be trusted as a personal assistant, as current systems continue to be illusory and biased.
But Lightcap says more advanced AI could at least take on the role of a “stand-up comedian,” and OpenAI executives say it could be possible.
“We've only scratched the surface of all the capabilities these systems have,” he said at a Milken Institute conference. “That would surprise us.”
Axel Springer, Business Insider's parent company, has a global deal that allows OpenAI to train models based on its media brands' reporting.
Axel Springer, Business Insider's parent company, has a global deal that allows OpenAI to train models based on its media brands' reporting.
