OpenAI has introduced new models, including an updated GPT-4 Turbo preview model, and has decreased the price of GPT-3.5 Turbo API access. The company is also launching features to assist developers in managing API keys and gaining insights into API usage.
The enhanced GPT-4 Turbo aims to address the issue of "laziness," where the model may fail to complete tasks thoroughly. OpenAI has also released a new GPT-3.5 Turbo model, gpt-3.5-turbo-0125, and has reduced the price of GPT-3.5 Turbo to support scalability. The input price for the new model is reduced by 50% to $0.0005 per thousand tokens, while the output price is reduced by 25% to $0.0015 per thousand tokens.
In response to user complaints about ChatGPT rejecting tasks, OpenAI attributes the issue to a lack of updates to GPT-4. While users of GPT-4 with data available before September 2021 may still encounter laziness issues, GPT-4 Turbo has been trained on the latest information from April 2023.
OpenAI has introduced smaller AI models known as embeddings, which are sequences of numbers representing concepts in language or code. These embeddings enhance computers' understanding and processing of written text, acting as efficient translators that convert human language into a format understandable by computers.
Retrieval-augmented generation, another AI type, focuses on providing accurate and relevant responses by looking up information rather than generating answers from scratch. OpenAI has introduced two new models, "text-embedding-3-small" and "text-embedding-3-large," with varying capacities to efficiently retrieve and use information in applications.
OpenAI is also working on enabling GPT creators to monetize their personalized AI systems. Initially, compensation for US builders will be based on user participation in their GPT, and a GPT store will be introduced first to users of paid ChatGPT plans.
