OpenAI and Meta are on the cusp of launching advanced AI models capable of reasoning and planning, which are fundamental for achieving superhuman cognitive abilities in machines. Both tech giants are gearing up to release the latest versions of their large language models that underpin generative AI tools such as ChatGPT. Meta is set to roll out Llama 3 soon, while OpenAI, backed by Microsoft, plans to release what is likely to be called GPT-5.
Meta’s Vice President Joelle Pineau and OpenAI’s Chief Operating Officer Brad Lightcap have both emphasized their focus on evolving these models to not just simulate conversation but to genuinely reason, plan, and remember. According to Lightcap, the forthcoming GPT model will significantly improve at solving challenging cognitive tasks. AI will be given the ability to perform a sequence of related tasks and predict the consequences of its decisions.
At a recent London event, Meta’s Chief AI Scientist Yann LeCun criticized existing AI systems for their sequential yet thoughtless processing, which often results in errors. By improving reasoning, AI could systematically explore options and forecast the consequences of actions, thus significantly advancing machine intelligence.
LeCun also shared insights into AI "agents" that could autonomously manage tasks such as planning a trip from Paris to New York. By the way, Meta plans to incorporate its new AI model into WhatsApp and its smart glasses, customizing Llama 3 for different applications over the next months. Generally, LeCun projected a future where advanced AI assistants are central to our digital interactions, influencing our entire digital lifestyle.
The race to refine generative AI is accelerating, with major companies like Google also launching new models this year. These advancements are crucial for developing what researchers term "artificial general intelligence" – AI that mimics human cognition.