Source URL: https://slashdot.org/story/24/12/22/0333225/openais-next-big-ai-effort-gpt-5-is-behind-schedule-and-crazy-expensive
Source: Slashdot
Title: OpenAI’s Next Big AI Effort GPT-5 is Behind Schedule and Crazy Expensive
Feedly Summary:
AI Summary and Description: Yes
Summary: The article discusses the challenges OpenAI is facing with the development of GPT-5, highlighting delays, high costs, and the struggle to gather adequate quality data. The issues point to broader concerns in the AI industry about potential stagnation in AI advancements, coinciding with opinions from key figures like Ilya Sutskever regarding the future of data availability.
Detailed Description:
The text offers an in-depth look at the ongoing development of OpenAI’s next-generation model, GPT-5, following the release of GPT-4. Here are the key points and insights:
– **Project Delays and Financial Strain**: OpenAI is reportedly behind schedule on GPT-5, accumulating substantial costs estimated to be around half a billion dollars for computing resources alone. This has raised questions about the project’s feasibility and timelines.
– **Data Gathering Challenges**: OpenAI’s team believes they need more diverse and high-quality data than what is available from public sources on the internet. This has led to a controversial strategy of generating synthetic data, which involves hiring experts to create and explain new materials specifically for training the model.
– **Human-Generated Data**: The emphasis on having subject matter experts (like software engineers and theoretical physicists) provide explanations for their work signifies a shift towards enriching the training data, potentially enabling the model to learn more effectively.
– **Internal Struggles and Talent Loss**: The organization has experienced significant turnover, with numerous key personnel, including co-founders and lead researchers, leaving OpenAI. This internal turmoil is compounded by competitive pressure from rival firms attempting to attract top talent.
– **Concerns Over Industry Progress**: There is a growing sentiment among community experts that the AI field may be hitting a plateau regarding improvements. Notably, Ilya Sutskever’s remarks on the diminishing returns of data availability have sparked discussions about the sustainability of current AI training methodologies.
– **Long-Term Implications**: If the ability to gather and effectively utilize data is indeed waning, this could signal a need for a fundamental reevaluation of AI development strategies and business models across the industry.
The concerns raised in this report reflect critical issues that professionals in AI, cloud, and infrastructure security must consider, particularly regarding the long-term viability of AI projects, the significance of data in training models, and the strategic management of human resources in talent-heavy sectors.