OpenAI’s GPT-4 Turbo Upgraded, Incomplete Task Execution & ‘Laziness’ Complaints Now Resolved

OpenAI has recently unveiled another innovation to its GPT-4 Turbo. The new innovation will be a significant upgrade aimed at excelling in tasks such as code generation while directly addressing the concerns that users raised regarding laziness and incomplete task execution. This comes in response to the recent complaints that alleged the predecessor of being lazy in completing prompted tasks.

OpenAI in a blog post informed about these proactive steps to enhance performance.

GPT-4 Vs GPT-4 Turbo 

Following the discontent among ChatGPT users regarding the refusal to complete tasks and the lack of updates in GPT-4, prompted OpenAI to clarify that the said update is just for GPT-4 Turbo. It said that this advanced version was trained on information as recent as April 2023 and it is currently available in a preview version.

While clarifying it said that users of the standard GPT-4, who are still relying on data up to September 2021, might experience the previously identified issues related to ‘laziness’ even now.

As per the blog post, over 70 per cent of users employing GPT-4 via its API have already shifted to GPT-4 Turbo owing to its more updated knowledge base. OpenAI has assured users that further updates for GPT-4 Turbo are also in the pipeline.

The enhancement in the capability due to this update will enable users to engage in much more complex multimodal prompts, for example, text-to-image generation.

ALSO READ | AI Has Been Somewhat Demystified Because People Really Use It Now: OpenAI Chief Sam Altman

More Updates: Embedding

In addition to this update, OpenAI also introduced smaller AI models known as embeddings in the blog post. These embeddings will play a crucial role in applications that are using retrieval augmented generation. This new and innovative approach will involve taking information from a database rather than generating answers independently.

OpenAI has launched two new models, text-embedding-3-small and the more potent text-embedding-3-large, which are now available for use.