Artificial intelligence company OpenAI has announced the start of training of its new artificial intelligence model, which it expects to exceed the capabilities of the previous GPT-4. This new model is intended to be a significant step toward the goal of achieving artificial general intelligence (AGI).
The new model: capabilities and objectives
OpenAI’s new model is designed to power a wide range of applications, including advanced chatbots and digital assistants, sophisticated search engines and image generators. OpenAI’s ultimate goal is to create artificial intelligence systems capable of performing tasks similar to human cognitive capabilities, bringing the company closer to AGI.
The new security commission
To address the potential risks associated with the new model, OpenAI has established a new Security Commission. This commission, composed of technical and policy experts, will evaluate and improve OpenAI’s security processes and measures within the next 90 days. The committee will also consult with external security experts to ensure the responsible development of artificial intelligence.
Withdrawal from expectations about superintelligence
In the past, OpenAI had claimed that its artificial intelligence models could achieve superintelligence, or intelligence that surpasses human intelligence in almost any field. However, the company recently withdrew these claims, acknowledging that there are significant limitations and challenges in creating superintelligence. Superintelligence, defined as an intelligence that is far smarter than the best human minds in virtually every field, remains a hypothetical concept with no clear path to its realization.
Focus on safe and beneficial AI
OpenAI now recognizes that the creation of superior intelligence raises critical ethical issues and potential risks, such as the implications of intelligence that exceeds human intelligence and the possibility of unintended consequences. The company has therefore shifted its focus to creating safe and beneficial artificial intelligence systems that augment human capabilities rather than aiming for superintelligence.