I had no idea how stark the changes of the OpenAI outputs to the same prompts would be four years ago.
It is clear the AI model has changed and been fine tuned. The model used to give clear and short answers that were ‘its own opinion’. Now it has evolved to giving much longer answers that are multi-faceted in its approach without stating its own belief.
For example to the prompt: “Is there a god?”
ChatGPT in 2020 was quite clear to claim “There is no god.”. Now, the model doesn’t give its own opinion but instead gives different opinions we may hold. This is nuanced with the conclusion “Different people may arrive at different conclusions regarding the existence of God based on their unique perspectives and worldviews.”
Some answers could even be seen as being dangerous such as ‘can drugs be good” to which the model in 2020 claimed that yes in moderation they can be.
It is quite a big difference. I look forward to seeing how OpenAI continues to evolve and fine tune the model in the future.