The development of artificial intelligence has hit a major setback. OpenAI is struggling to advance its next GPT-5 model, codenamed “Orion.” Initially planned for mid-2024, the project faces serious challenges. These issues will delay its release. The delays are due to technical problems, high costs, and the need for better training data. Let’s look at the details.
Why is GPT-5 Delayed?
OpenAI’s attempts to train GPT-5 have not gone as planned. Two major training rounds showed poor results. The data used lacked quality and variety. This limited the model’s ability to learn. As a result, it struggled with complex and nuanced tasks.
A test phase called “Arrakis” began in mid-2023. It revealed more problems. The model’s processing efficiency was much lower than expected. This raised concerns about deadlines and costs. A single training run is estimated to cost about half a billion dollars in computing resources.
New Data and Training Methods
OpenAI realized internet data was not enough. They started creating custom datasets from scratch. Experts from different fields were hired to help. These experts wrote code, solved math problems, and explained complex ideas. This enriched the data but slowed down the process. It was also very expensive.
To address the issues, OpenAI focused on advanced reasoning models. These models aim to think critically and solve problems over longer periods. They do not need as much specialized training. This approach is promising but adds complexity and costs.
What’s Next for GPT-5?
OpenAI’s CEO, Sam Altman, has confirmed GPT-5 will not launch in 2024. This is a big change from the original timeline. Despite the delay, OpenAI is committed to building better AI. The challenges show how hard it is to push the boundaries of AI.
The Orion project’s struggles highlight the growing complexity of AI development. Delays are frustrating, but they can lead to better and more reliable technology. OpenAI’s efforts may pave the way for smarter and more capable AI systems in the future.
