If true, this would be a significant step forward in creating machine cognition, and further developing what is known as “Artificial General Intelligence” – AGI – which would help robots equal or surpass human intelligence.
Illustration: FT
Executives at OpenAI and Meta both signaled this week that they are preparing to release next versions of their large language models — systems that serve as the foundation for building AI applications like ChatGPT.
Meta says it will begin rolling out Llama 3 in the coming weeks, while Microsoft-backed OpenAI indicates that its next model, expected to be called GPT-5, will be “coming soon.”
“We're trying to figure out how to not only make these models talk, but also reason, plan… and have memory,” said Joelle Pineau, vice president of AI research at Meta.
An OpenAI director, Brad Lightcap, told the Financial Times that the next generation of GPTs will show progress in solving “hard problems” like reasoning.
“We’re going to start seeing AI take on more complex tasks,” he said in an interview. “I think we’re just scratching the surface of what these models can do to reason.”
The Meta and OpenAI upgrades are part of a wave of new big language models being released this year by companies including Google, Anthropic, and Cohere.
Tech companies are racing every day to create ever more complex AI software, from being able to generate text, images, write code and create videos like humans.
Reasoning and planning are important steps toward what AI researchers call “Artificial General Intelligence,” or AGI, which could help machines achieve human-level cognition or intelligence. AGI will also allow chatbots and virtual assistants to complete sequences of tasks and predict the consequences of their actions.
Hai Anh (according to FT)
Source
Comment (0)