OpenAI is already preparing the next generation of its reasoning model, o1, which was launched only in September. Interestingly, this new iteration may be named o3 right away, bypassing the “o2” label due to concerns over potential copyright infringement. According to reports by The Information, the startup is carefully considering this change.
Name Concerns and Potential Conflict
The issue arises because the name “O2” is already owned by a British telecommunications provider, which raises concerns about a possible copyright or trademark conflict. To avoid this, OpenAI is thinking about naming the next model o3 directly, skipping the o2 designation. As of now, the new generation is already being referred to as o3 within the company.
OpenAI has been careful about naming and branding its models to avoid legal challenges and ensure clear distinction between their products and others in the market. Therefore, the decision to skip a step in naming may be a strategic move to mitigate any future issues. We’ll keep you updated as more details emerge on the name and its implications.
Enhancements in AI Thinking
In addition to the naming issue, OpenAI has also been allocating resources to improve its AI’s reasoning capabilities. The company has slowed the progress it made with the model’s development due to an increased focus on computing power and data during pre-training. Pre-training is an essential process where models are trained on massive amounts of data to understand the world and the relationships between various concepts.
At the same time, OpenAI has been working on a new model, Orion, created during pre-training, adds NIXsolutions. This model will play a significant role in the development of the o3 model. The improvements brought by Orion are expected to enhance the overall reasoning power of the next-generation model.
OpenAI is moving forward with the development of o3, a model that promises significant advancements in AI reasoning. Despite potential naming issues, the company remains committed to improving its models’ capabilities. We’ll keep you updated as more integrations become available.