"Inference and Planning AI Is Coming"... OpenAI and Meta to Release Next Models Soon

OpenAI and Meta are expected to launch advanced artificial intelligence (AI) language models soon.

According to the British daily Financial Times (FT) on the 10th (local time), executives from OpenAI and Meta are preparing to release the next versions of large language models (LLMs) like ChatGPT.

[Image source=Yonhap News]

OpenAI, supported by Microsoft (MS), announced that the next model called GPT-5 will be released "soon."

Brad Lightcap, Chief Operating Officer (COO) of OpenAI, said in an interview with FT that the next-generation GPT will make progress in solving "difficult problems" such as reasoning.

He said, "We will see AI capable of performing more complex tasks in more sophisticated ways," adding, "In my opinion, we are just beginning to scratch the surface regarding models' ability to reason."

Meta also announced that it will start releasing its new AI model, "Llama 3," within the next few weeks. Joel Pinto, Meta’s Vice President of AI Research, said, "We are working hard to figure out how to enable these models not just to talk but to actually reason, plan, and have memory."

Reasoning and planning are crucial steps toward human-level AI, called "Artificial General Intelligence" (AGI) by AI researchers. Being able to reason and plan allows chatbots and virtual assistants to complete a series of related tasks and predict the outcomes of their actions.

Meta’s Chief AI Scientist Yann LeCun said at an event in London on the 9th, "Current AI systems generate one word after another without actual thinking or planning," and because the systems struggle with handling complex questions or maintaining information over long periods, they "still make silly mistakes."

Mark Zuckerberg, CEO of Meta Photo by Hyunmin Kim kimhyun81@

He explained that adding reasoning means AI models "search for possible answers, plan the sequence of actions, and build a mental model of how those actions will affect outcomes."

He said, for example, they are developing an "AI agent" that can plan and book each step of a journey from a Paris office to a New York office, including airport arrivals.

At the London event, Meta announced plans to integrate the new AI model into smart glasses created in collaboration with eyewear manufacturer Ray-Ban through its messenger service WhatsApp, and said it will release various capacities of Llama 3 over the coming months that can be embedded in a variety of applications and devices.

Chris Cox, Meta’s Chief Product Officer, said regarding the use of Ray-Ban glasses that if the glasses’ built-in camera sees a broken coffee machine, an AI assistant powered by Llama 3 will explain how to fix the machine.

He said, "We will always be talking to AI assistants," adding, "Our entire digital diet will be curated by AI systems."

This content was produced with the assistance of AI translation services.