تخطط OpenAI لإطلاق نموذج الذكاء الاصطناعي Orion: أقوى 100 مرة من GPT-4

OpenAI's Orion: A 100x More Powerful AI Model Than GPT-4

OpenAI is gearing up to unveil its next-generation AI model, Orion, in December 2024. This groundbreaking model will succeed the current ChatGPT-4, but unlike its predecessors - GPT-4o and o1 - it won't be readily available to the public through ChatGPT.

Instead, OpenAI plans to grant initial access to select companies it collaborates with directly. These businesses will be able to leverage Orion's capabilities to develop their own products and features. Microsoft engineers, OpenAI's primary partner, are reportedly preparing to host Orion on Azure as early as November, making the platform accessible to companies before the general public.

While Orion is considered a new iteration of GPT-4, its official public release name remains uncertain, leaving open the possibility of being labeled as GPT-5. Notably, an OpenAI executive revealed that Orion boasts a power level up to 100 times greater than GPT-4, pushing the boundaries towards achieving artificial general intelligence (AGI).

This advanced model is being developed independently of OpenAI's recently launched o1 reasoning model. Orion's training process leverages synthetic data generated by o1, internally codenamed "Strawberry."

Orion's arrival coincides with a pivotal time for OpenAI, which recently secured $6.6 billion in funding, prompting the company to restructure as a profit-seeking entity. The company has also witnessed significant personnel changes, including the resignation of Chief Technology Officer Mira Murati, Head of Research Bob McGrew, and Vice President of Post-Training Barrett Zoph.

This timeline aligns with a cryptic X post by OpenAI CEO Sam Altman, hinting at "winter constellations" soon to emerge. When queried about Altman's post, ChatGPT o1 - preview suggests it's an allusion to the term "Orion"!