OpenAI and Meta Poised to Produce AI Models Capable of Reasoning and Planning
According to a report by the Financial Times, tech giants OpenAI and Meta are on the cusp of introducing advanced AI models poised to redefine the landscape of machine cognition. These next-generation models are not only expected to enhance conversational AI applications like ChatGPT but are also designed to master reasoning and planning.
If proven successful, these can become the cornerstones of what experts term as artificial general intelligence, or AGI. So what’s going on? Well, Meta is set to unveil Llama 3, its latest iteration in the coming weeks.
At the same time, OpenAI, with Microsoft’s backing, hints at the imminent launch of what is likely to be dubbed GPT-5. All of this together is fueling a grounds swell of speculation of models that could turn the page on AI’s next chapter.
Joelle Pineau, Vice President of AI Research at Meta, emphasized the transformative capabilities of the upcoming models, highlighting their ability to reason, plan, and remember. It is expected to bridge the gap between current AI’s proficiency in executing isolated tasks and its aspiration for comprehensive cognitive functions.
Echoing this sentiment, Brad Lightcap, COO of OpenAI, shared with the Financial Times their vision for the next-gen GPT, focusing on its enhanced problem-solving prowess. “We’re going to start to see AI that can take on more complex tasks in a more sophisticated way,” Lightcap commented.
As many know, the push for reasoning and planning capabilities is seen as a crucial step towards achieving AGI, to enable virtual assistants and chatbots to execute a sequence of interconnected tasks and predict the outcomes of their actions.
Yann LeCun, Meta’s Chief AI Scientist, critiqued current AI systems for their lack of foresight and planning, pointing out the necessity for AI to construct a “mental model” to anticipate the consequences of its actions, thereby significantly reducing errors.
The forthcoming models from Meta and OpenAI are part of a broader trend of significant advancements in generative AI, alongside contributions from Google, Anthropic, and Cohere. These developments are set to accelerate the pace at which AI can produce human-like text, images, code, and videos, pushing the boundaries of creativity and efficiency.
Also, Meta plans to integrate AI assistants across its platforms, including WhatsApp and its Ray-Ban smart glasses, showcasing practical applications that could transform everyday tasks. Chris Cox, Meta’s Chief Product Officer, illustrated a future where AI assistants, powered by Llama 3, provide real-time guidance on tasks as mundane yet complex as fixing a coffee machine.
