Proposal for AI Translation with Interactive Learning to Improve Context and Precision

I want to share an idea that I think could significantly improve current machine translation tools. Although translation models have come a long way, they still often make mistakes in specific contexts, such as translating ambiguous words without taking into account the correct meaning of the context.

My proposal is to develop a translation model with an interactive learning approach. The goal would be to allow AI to learn like a human learner, correcting its errors step by step and retaining those corrections to apply them in the future. It doesn’t have to be created from scratch, just use a translation model like Google’s and correct the errors.

Once the “teaching” of AI, which until now was like a chatGPT type model, is finished, it is transformed into a simple script or tool with which the only interaction is to write the text to be translated, thus turning it into something light in weight and with almost no server maintenance costs.
I must say that this has been written with a translator because I do not know English, so I apologize for any errors in writing (and this is one of the reasons why I propose this).
I also apologize if this post is in the wrong section, I’m not sure which one it should be in.