- OpenAI and Meta are close to unveiling AI models that can reason and plan, the FT reported.
- An OpenAI exec told the FT that the next version of GPT would show progress with "hard problems."
- Some experts warn that developing such technology could put humans at risk of extinction.
OpenAI and Meta are reportedly preparing to release more advanced AI models that would be able to help problem-solve and take on more complex tasks.
OpenAI's chief operating officer, Brad Lightcap, told The Financial Times that the company's next version of GPT would show progress on solving "hard problems," such as reasoning.
"I think we're just starting to scratch the surface on the ability that these models have to reason," he told the outlet.
Meta's executives signaled a similar trajectory for the company's upcoming Llama 3 model, expected in the coming weeks.
Joelle Pineau, vice-president of AI research at Meta, said the company was "hard at work" figuring out how to get the model to talk, reason, plan, and have memory.
Representatives for Meta and OpenAI did not immediately respond to a request for comment from Business Insider, made outside normal working hours.
Getting AI models to reason and plan is an important step toward achieving artificial general intelligence (AGI), which both Meta and OpenAI have claimed to be aiming for.
The development could be worth trillions for the company that achieves it. In February, ex-Meta executive and virtual reality visionary, John Carmack, called artificial general intelligence (AGI) AI's "big brass ring." He claimed it would become a trillion-dollar industry by the 2030s.
While the definitions of AGI can vary, it is most simply defined as a type of artificial intelligence that can perform at a human level or better on a wide range of tasks.
Some experts have raised safety concerns about developing technology that exceeds human intelligence. Notable researchers, including AI godfathers Yoshua Bengio and Geoffrey Hinton, have urged people to consider the risks to humanity from AI.
Elon Musk, a longtime AI skeptic, recently estimated that AI would outsmart humans within two years. Musk said the "total amount of sentient compute" — a concept that may refer to AI thinking and acting independently — will exceed all humans in five years.