Alphabet Inc. Class C, Anthropic, and OpenAI are all stuck in a "deadlock" in AI model development. How to break through the bottleneck?

According to sources familiar with the matter, like OpenAI, Alphabet Inc. Class C (GOOGL.US) and Amazon.com, Inc. (AMZN.US) supported Anthropic's large language model performance has also entered a stagnant bottleneck period. Three sources said that the upcoming Gemini did not meet internal expectations. Additionally, Anthropic has delayed the release of the next version of its Claude model, 3.5 Opus. The companies stated that there is currently a lack of artificial data available to train the models. At a critical milestone moment, OpenAI completed the first round of training for a large new artificial intelligence model in September, hoping that the model would greatly surpass the previous technological versions behind ChatGPT and bring the goal of surpassing human-level artificial intelligence closer. However, according to two sources, the model internally known as "Orion" did not meet the company's expected performance. Sources said that, for example, by the end of the summer, Orion performed poorly when attempting to answer untrained programming questions. Overall, so far, the improvement of Orion over OpenAI's existing models has not been as significant as the upgrade from GPT-3.5 to GPT-4. GPT-3.5 was originally the flagship chat system for OpenAI's Siasun Robot & Automation. OpenAI is not the only company facing obstacles recently. After launching increasingly complex artificial intelligence products at a rapid pace for many years, three leading artificial intelligence giants are currently finding that the diminishing returns from the high costs of developing new models. Anthropic released a new version called Claude 3.5 Haiku last month, as well as an upgrade version of Claude 3.5 Sonnet. In September of this year, OpenAI introduced a series of new artificial intelligence models, named o1 and o1-mini, capable of reasoning complex tasks and solving more difficult problems in science, programming, and math than previous models. However, according to three sources, the upcoming Gemini software version from Alphabet Inc. Class C did not meet internal expectations. Meanwhile, the long-awaited release schedule for Anthropic's Claude model 3.5 Opus has also been delayed and no specific release date has been provided. These AI giants are currently facing some challenges. Finding new, undeveloped sources of high-quality artificial training data to build more advanced artificial intelligence systems is becoming increasingly difficult. Two sources said that Orion's programming performance is not ideal, in part due to a lack of sufficient programming data for training. Meanwhile, even moderate improvements may not be enough to justify the huge costs associated with creating and operating new models, as well as meeting the expectations of marketing a product as a significant upgrade. However, these models still have great potential for improvement. According to one source, OpenAI has been putting Orion through a process called "fine-tuning" for several months. This is a regular process before the company publicly releases new artificial intelligence software, including incorporating human feedback to improve responses and improving the tone of the model's interactions with users. But one source said that Orion has not yet reached the level that OpenAI hopes to release to users, and the company is unlikely to launch the system before early next year. The Information previously reported some details of the challenges OpenAI is facing in developing new models, including programming tasks. Furthermore, OpenAI CEO Sam Altman previously stated that overall computational power is also an issue that the latest wave of artificial intelligence models needs to address. Altman recently said on Reddit, "All these models are getting quite complex, and we can't release so many things all at once. We also face a lot of constraints and tough decisions on how to allocate our compute to many great ideas."