Beyond Chatbots: The Dawn of AI and Reasoning
OpenAI and Meta are on the verge of unveiling advanced AI models designed to enhance reasoning and planning abilities, a significant step towards achieving unparalleled machine intelligence.
The Daily Dose
The Daily Dose Quiz is a fun newsletter that offers quizzes covering topics like words, numbers, and general knowledge. It’s updated regularly, with new issues being published frequently. Subscribe and learn more here: The Daily Dose
OpenAI and Meta are on the verge of unveiling advanced AI models designed to enhance reasoning and planning abilities, a significant step towards achieving unparalleled machine intelligence.
Leaders from both OpenAI and Meta have announced their plans to introduce the latest iterations of their large language models, the foundation for generative AI tools like ChatGPT.
Meta has announced the upcoming launch of Llama 3, and OpenAI, supported by Microsoft, has hinted that its forthcoming model, likely to be named GPT-5, will be available shortly.
Joelle Pineau, Meta's vice-president of AI research has emphasized the goal of evolving these models beyond simple conversation to include reasoning, planning, and memory capabilities.
Brad Lightcap, OpenAI's COO, revealed that the new generation of GPT would advance in tackling complex problems, particularly in reasoning. While current AI can handle basic tasks well, its capabilities remain quite limited. Lightcap anticipates that the upcoming models will handle more complex tasks with enhanced sophistication, barely scratching the surface of their potential to reason.
This development is part of a broader movement, with companies like Google, Anthropic, and Cohere also releasing new large language models this year. Advancements in reasoning and planning are crucial for achieving "artificial general intelligence," or machine-based cognition at human levels, enabling more complex task sequences and predictive actions.
At an event in London, Yann LeCun, Meta's chief AI scientist, criticized current AI systems for their lack of foresight and tendency to make errors due to an inability to think ahead or retain information over long periods.
LeCun explained that adding reasoning to AI would involve exploring possible answers, planning action sequences, and understanding the outcomes of these actions, a vital step towards elevating machine intelligence.
Meta is also working on AI "agents" capable of planning and managing complex activities, such as arranging a trip from Paris to New York, and plans to integrate the new Llama 3 model into WhatsApp and Ray-Ban smart glasses.
Lightcap stated that more details on the next GPT version would be disclosed soon, emphasizing the shift towards handling longer and more complex tasks that necessitate improved reasoning abilities.
At Meta's London event, Chris Cox, the chief product officer, suggested that Meta's Ray-Ban glasses, equipped with cameras and AI assistance from Llama 3, could offer practical solutions, such as fixing a broken coffee machine.