🤔 Did you know that ChatGPT has been "silently degraded" over time?
A Stanford University study found that the behavior of GPT-3.5 and GPT-4 has varied significantly - there are side-effects as these models are fine-tuned with additional data.
Performance and accuracy have decreased over time for some tasks (while improved for others).
If you have any systems relying on those models, you better have automated tests and keep monitoring.
🔗 LINK: https://lnkd.in/g_xu8vWe
