The end of GPT-4? The chief of OpenAI reveals the future innovation direction of AI
According to the latest reports, OpenAI CEO Altman recently issued a striking statement. He believes that we have entered a new era of artificial intelligence research, in which the strategy of improving performance by simply increasing the size of the model is outdated. This is an important observation because this strategy has driven many major advances in AI over the past few years, including OpenAI's own GPT-3 and GPT-4 models.
Altman's point is that future AI advances will require more complex strategies than just increasing model size. He pointed out that although GPT-4 is a very powerful model, its performance improvement is not linear, that is, the performance cannot be continuously improved simply by increasing the size of the model. Additionally, he noted that physical constraints, such as the number of data centers and how quickly they can be built, also limit model size growth.
That's not to say that GPT-4 or similar large models don't have value. In fact, GPT-4 has shown impressive performance in many applications, including in Microsoft's Bing search engine and Google's Bard chatbot. However, Altman's view is that future AI research will require more sophisticated strategies, including new AI model designs or architectures, and further adjustments based on human feedback.
This insight is an important reminder for AI researchers and practitioners that we cannot rely solely on increasing model size to drive AI progress. We need to find new strategies and methods to overcome the limitations of existing technologies and promote the further development of AI technology. This could include new model architectures, more efficient training methods, or more complex AI system designs.
Overall, Altman's announcement is an important milestone, marking our entry into a new era of AI research. In this new era, we will need more sophisticated strategies and methods to drive AI progress beyond just increasing model size. This is an important reminder for AI researchers and practitioners, and it also provides a new direction for future AI research.
