Recent reports indicate that OpenAI, Google, and Anthropic are encountering significant challenges in advancing their next-generation AI models. Despite the excitement surrounding AI upgrades, the anticipated enhancements may not materialize as quickly as hoped.
OpenAI's Orion model, for instance, is reportedly not meeting performance expectations, particularly in coding tasks. This has led CEO Sam Altman to downplay rumors regarding its release date and upgrades to ChatGPT. Similarly, Google's Gemini project is progressing slower than anticipated, while Anthropic has delayed the launch of its Claude 3.5 Opus model.
All three companies are facing limitations related to training data. While vast datasets have been utilized, the availability of high-quality, unused information is dwindling. This scarcity, combined with ethical considerations surrounding data usage, poses a challenge to further improvements in AI capabilities. Even with sufficient raw data, the costs associated with processing and integrating it into AI models are substantial.
To navigate these hurdles, developers are exploring alternative enhancement methods, such as incorporating human feedback post-training. However, this approach is slower and raises concerns about the potential limits of scaling AI models effectively.
As a result, users may experience a gradual release of new AI features. This pace could provide an opportunity to fully explore and utilize existing AI tools, such as ChatGPT-o1. Additionally, this slowdown might allow OpenAI to focus on launching its long-anticipated Sora AI video creator, which has remained under wraps despite ongoing teasers.