OpenAI, Meta to Roll Out AI Versions Capable of Planning, Reasoning

Microsoft-backed OpenAI and Meta will soon release artificial intelligence (AI) models capable of reasoning and planning, according to a report.

OpenAI, Meta

Meta has announced plans to release Llama 3 in the coming weeks, while OpenAI has revealed that its forthcoming AI model ‘GPT-5’, which will also be introduced soon, the report added.

According to the FT report, Joelle Pineau, vice-president of AI research at Meta, said that the company is actively exploring the capabilities of large language models to engage in reasoning, planning, and memory functions. He said these models, which are the backbone of popular generative AI applications such as ChatGPT and Llama, are being studied for further enhancements.

Brad Lightcap, chief operating officer (COO) of OpenAI, said the upcoming version of GPT is aimed to tackle challenging tasks, including reasoning, the FT reported.

Lightcap said the current AI systems only have limited capabilities and are merely scratching the surface. However, he said, AI will evolve to handle more intricate tasks in a sophisticated manner.

The competition in the field of AI is intensifying as rival companies such as Google and Anthropic, prepare for new launches. Meta plans to incorporate its latest AI model into WhatsApp and Ray-Ban smart glasses, the FT reported. Additionally, the company is preparing to introduce Llama 3 in different sizes, customised for various applications and devices, over the next few months.

OpenAI launches GPT-4 Turbo

On April 10, OpenAI launched an upgraded edition of GPT-4 Turbo, which is now available via an application programming interface (API) and accessible to subscribers of ChatGPT Plus. This latest version, GPT-4 Turbo, signifies the company’s most recent advancement in generative AI models, having been trained with data up to December 2023.

According to a report in the Indian Express, the revised model features enhanced data processing capabilities that allow for text analysis in JSON format and the execution of function calls to produce JSON code snippets. These snippets help developers to automate tasks within their integrated applications, such as sending emails, posting online content, or completing purchases.