OpenAI Unveils Revolutionary New Artificial Intelligence Technology

On July 15, reports from Russia Today highlighted a recent article from Reuters, which revealed that OpenAI, the creator of the ChatGPT technology, is developing a new form of artificial intelligence (AI). As part of the project codenamed "Strawberry," this Microsoft-backed organization aims to significantly enhance the reasoning capabilities of its models. An insider shared that the workings of "Strawberry" are kept under strict confidentiality, even within OpenAI.

According to the source, the project involves a specialized approach to training an AI model that has already undergone extensive pre-training on large datasets. The goal is to enable the AI not only to answer questions but also to proactively gather comprehensive information through autonomous and reliable internet browsing, facilitating what is termed as "deep research." Reuters noted that it had accessed an internal OpenAI document outlining research plans linked to "Strawberry." However, it remains unclear when this technology will be available to the public.

In response to inquiries, an OpenAI spokesperson stated, “We hope our AI models will understand and perceive the world as humans do. Continuously researching new AI capabilities is standard practice in the industry, fueled by a shared belief that these systems will enhance their reasoning abilities over time.” This spokesperson did not directly reference the "Strawberry" project.

Currently, large language models can summarize vast amounts of text and produce coherent articles more rapidly than humans, yet they often struggle with straightforward, commonsense reasoning. In such cases, AI models can "hallucinate," presenting false or misleading information as fact. Researchers interviewed by Reuters emphasized that improving reasoning is crucial for AI to reach or surpass human-level performance, which current models have not yet achieved.

Last week, renowned AI expert and deep learning pioneer Yoshua Bengio warned about the risks associated with corporations racing to develop AI that matches or exceeds human capabilities, including the potential threat to humanity. Bengio urged the scientific community and society at large to engage in a "collective effort" to find ways to control advanced AI technologies.