OpenAI announced that the bug where ChatGPT was lazy and did not complete the coding tasks assigned to it was fixed. However, the update in question has only been made to GPT-4 Turbo.
Recently, a large number of ChatGPT users were complaining about the AI-powered chatbot frequently refusing to complete requested tasks. OpenAIIn a statement shared last night, , announced that it has updated ChatGPT for related problems, but the update is currently only for GPT-4 Turbo. Therefore, standard ChatGPT users may continue to see related issues.
OpenAI did not disclose what exactly it changed in GPT. However, in the statement made on OpenAI Developers’ X account, “The updated GPT-4 Turbo preview more comprehensively completes tasks like encoding and fixes a bug in UTF-8 encoding” expressions were used. Also, starting next week ChatGPT-3.5 TurboIt was stated that input prices would be reduced by 50% and output prices would be reduced by 25%.
Fixed bug where GPT-4 Turbo was rejecting missions
OpenAI also states that GPT-4 Turbo will be available for general availability as a full version in a few months. This is extremely important because GPT-4 has a relatively primitive infrastructure that has “laziness” issues such as not performing some tasks trained on pre-September 2021 data, while it is still available as a preview. GPT-4 TurboIt was trained with a much more up-to-date data package, such as April 2023, and the “laziness” problems in question have also been resolved.
The highlights of OpenAI’s new GPT-4 Turbo update are as follows:
- New docking models with lower prices
- A new small text placement model
- Native support for shortening embeds
- Other new models and price reductions
- GPT-3.5 Turbo model updated and price reduced
- GPT-4 Turbo preview updated
- Control model updated
You can access OpenAI’s latest update notes here.