According to a report from The Information, the next OpenAI’s flagship model, code-named Orion, delivers improved performance over the other products. Despite this, the advancements are less substantial compared to the shift from GPT 3 to GPT 4. This insight came from employees who have already tested the upcoming model.
Reportedly, Orion might not outperform other Open AI models in areas such as coding. That’s why the latest OpenAI news suggests that the rate of improvement on the company’s latest product is slowing down.
However, this is a general issue for most LLM (large language model) products. The emerging pattern is because each generation shows less progress in terms of power and range. With the GPT 3, its responses to prompts were usually unpredictable. The GPT 3.5 improved on this, but the responses were still inconsistent at times. But then, the GPT 4 was released, and it was the first chatbot even to show a certain level of reasoning, followed by its human-like chatbot.
But will GPT 5 be able to outperform this giant leap? Although it is highly anticipated, the future remains uncertain with the recent slowdown in AI progress.
This is why OpenAI has developed a team to understand how the company can improve its models. This new strategy has reportedly involved training the new model, Open AI Orion, using synthetic data generated by other AI models. At the same time, they are also focusing more on the post-training process – a step that the previous models lacked.
While OpenAI news did not deny or agree with the issue that the company is reportedly facing, this might be the reason why they are postponing the release of its next model. Although GPT 5 was set to be launched by December 2024, OpenAI told TechCrunch: “We don’t have plans to release a model code-named Orion this year,
At the same time, it seems that the number of people using ChatGPT is slowly decreasing. This situation raises questions regarding the growth of OpenAI’s products, prompting the company to reassess its strategy.
This slowdown came especially after the ChatGPT down incident on November 8th, which affected hundreds, if not thousands, of users. But there’s no need to panic! "Issues with ChatGPT have now been resolved," stated OpenAI.
.webp)
.webp)
