OpenAI's ChatGPT Team Predicts Emergence of Superintelligent AI Within a Decade: Researching Ways for AI to Oversee AI

Recently, we explored the enhanced capabilities of ChatGPT following the release of the GPT-4 language model, which now features plugins such as a web browser and a code interpreter. These advancements significantly improve ChatGPT's ability to utilize tools, connect online, and perform complex computations. Global media feedback indicates that GPT-4 has surpassed GPT-3.5 in performance. However, the rapid advancements have raised concerns regarding personal information security and user control over ChatGPT.

In a noteworthy development, Meta's CEO Mark Zuckerberg and OpenAI's CEO Sam Altman have both expressed their support for AI regulation initiatives proposed by EU Commissioner Thierry Breton. During a recent meeting, they agreed on the necessity of government oversight to assess AI risks and regulate the technology's development. Altman praised the EU's commitment to regulating AI and showed eagerness to collaborate with Breton to address local market needs.

Breton is also engaging with other industry leaders, including NVIDIA CEO Jensen Huang and Twitter's new CEO Linda Yaccarino, to promote AI regulation. Following these discussions, OpenAI announced plans to allocate more resources and form a new research team dedicated to ensuring AI safety, aiming to develop systems that can effectively oversee other AI technologies.

OpenAI co-founder Ilya Sutskever and AI alignment head Jan Leike emphasized in a blog post the potential dangers posed by superintelligent AI, warning that its powerful capabilities could threaten humanity's survival. They stressed the urgent need for stronger technologies to manage superintelligent AI and called for significant advancements in AI alignment research to ensure that AI goals remain aligned with human values.

To address these concerns, OpenAI plans to dedicate 20% of its computing power over the next four years to AI alignment research. A new "Super Alignment" team will focus on developing AI-driven alignment researchers at a “human level,” utilizing substantial computational resources for their work. OpenAI also aims to incorporate human feedback into its training and evaluation processes for AI systems.

However, AI safety advocate Connor Leahy has criticized OpenAI's approach, highlighting potential flaws. He argues that early human-level AI could become uncontrollable and harmful, underscoring the need for immediate focus on AI safety. "We must solve the alignment problem before developing human-level intelligence, or we risk losing control," he stated.

As global concerns about AI safety intensify, rapid developments and measures are emerging. Staying informed on this critical topic is essential for those invested in the future of AI.