OpenAI cures GPT-4 ‘laziness’ with new updates

A rendition of OpenAI’s logo, which looks like a stylized whirlpool.
Illustration: The Verge

In a blog post, OpenAI said the updated GPT-4 Turbo “completes tasks like code generation more thoroughly than the previous preview model and is intended to reduce cases of ‘laziness’ where the model doesn’t complete a task.”

The company, however, did not explain what it updated.

Some ChatGPT users recently complained about the chatbot frequently refusing to complete prompted tasks and blamed the lack of updates to GPT-4. However, OpenAI’s update is for GPT-4 Turbo, a version of the more widely available GPT-4 that was trained on information as recent as April 2023 and is only available in a preview. Those using GPT-4, which learned from data available prior to September 2021, may still experience the same laziness issues.

OpenAI said in the post that more than 70 percent of those using GPT-4 via its API have moved to GPT-4 Turbo because of its more updated knowledge base.

The company said more updates to GPT-4 Turbo are coming in the next few months, including the general availability of GPT-4 Turbo with vision. This will allow users to do more multimodal prompts like text-to-image generation.

OpenAI also launched smaller AI models called embeddings. OpenAI defines Embeddings as a “sequence of numbers that represents the concepts within content such as natural language or codes.” This helps applications using retrieval-augmented generation — a type of AI that takes information from a database instead of generating its answer — figure out the relationship of different contents it’s accessing.

These new models, text-embedding-3-small and a more powerful version called text-embedding-3-large, are available now.