OpenAI and Meta are gearing up to introduce new artificial intelligence models designed to enhance reasoning and planning capabilities, marking significant strides towards achieving superhuman cognition in machines.
This week, executives from both OpenAI and Meta hinted at imminent launches of their latest iterations of large language models, integral to powering generative AI applications like ChatGPT.
Meta announced the impending rollout of Llama 3 in the upcoming weeks, while OpenAI, backed by Microsoft, teased the release of its anticipated GPT-5 model in the near future.
Joelle Pineau, Meta’s vice-president of AI research, emphasized the ongoing efforts to imbue these models with the capacity for reasoning, planning, and memory retention, going beyond mere conversational abilities.
Brad Lightcap, Chief Operating Officer at OpenAI, underscored the forthcoming advancements in solving intricate problems such as reasoning, heralding a new era where AI can tackle more complex tasks with heightened sophistication.
While current AI systems excel at discrete tasks, Lightcap noted their limitations in broader contexts, signaling the potential for enhanced reasoning capabilities to expand their functionality significantly.
Meta and OpenAI’s initiatives are part of a broader trend in the tech industry, with companies like Google, Anthropic, and Cohere also releasing upgraded large language models this year, accelerating the pace of progress in generative AI.
Reasoning and planning represent crucial milestones in the pursuit of artificial general intelligence, enabling chatbots and virtual assistants to navigate sequences of tasks and anticipate outcomes—a key step towards achieving human-level cognition.
During an event in London, Meta’s chief AI scientist Yann LeCun emphasized the current limitations of AI systems, highlighting their inability to think and plan comprehensively, resulting in errors and oversights.
Integrating reasoning capabilities would enable AI models to deliberate over potential solutions, strategize sequences of actions, and anticipate outcomes—a significant advancement in enhancing machine intelligence.
Meta plans to integrate its new AI model, Llama 3, into platforms like WhatsApp and its Ray-Ban smart glasses, catering to diverse applications and devices, with rollout plans spanning the coming months.
OpenAI is poised to unveil further details about GPT-5, with Lightcap hinting at enhanced capabilities for tackling more intricate tasks, underscoring the pivotal role of improved reasoning abilities in advancing AI capabilities.
At Meta’s event, Chief Product Officer Chris Cox showcased potential applications, envisioning AI-powered assistants guiding users through tasks like troubleshooting broken appliances, indicative of a future where AI mediates various aspects of daily life.