OpenAI simply dropped 4 main updates in a weblog submit at this time:
- An up to date GPT-4 Turbo preview mannequin
- An up to date GPT-3.5 Turbo mannequin
- An up to date textual content moderation mannequin
- Two new embedding fashions
For starters, GPT-3.5 is the bottom free language mannequin that powers the free model of ChatGPT. If you’re a paying subscriber, you’re probably utilizing GPT-4 language fashions (though you’ll be able to swap to GPT-3.5 for those who desire).
New GPT-4 Turbo Model
The new mannequin is known as “gpt-4–0125-preview.” It accommodates a repair for the bug impacting non-English UTF-8 generations. But what’s an much more fascinating replace to this mannequin is the repair for the difficulty of “laziness.”
According to OpenAI, the brand new mannequin “completes duties like code technology extra totally than the earlier preview mannequin and is meant to cut back circumstances of “laziness” the place the mannequin doesn’t full a process.”
What does that imply precisely?
In OpenAI’s group boards, some customers are complaining that ChatGPT typically refuses to execute duties that contain loads of information.
I’ve skilled the identical situation myself—ChatGPT will be fairly lazy at instances, refusing to offer detailed responses or generate lengthy texts. Have you observed this laziness as properly?
In a tweet on December 2023, OpenAI acknowledged that GPT-4 was getting lazier, however it wasn’t intentional.