Thank you. That is a good example.

I get that. But the game is “finished”. there is no need for alterations. translating the assembler code into c++ in this way could serve to quickly get it in a format that is then compileable for other platforms.

But do we need this information then? E.g. shouldn’t it be possible to just write what the assembler is doing as a c++ code?

E.g. high level languages also support stuff like bitwise operators and so on.

They could just have the kids read actual books designed by actual pedagogic experts which actually help to learn through studying it.

Now nobody knows if the “AI” is even teaching real things or if it is only using properly vetted material, if the structure it proposes makes sense.

Yes teachers are fallible, but they are also human and can emotionally understand what is going on during learning that a trained algorithm just cannot get. In so far also it means there needs to be a clearly defined “goal” of knowledge and competencies and the algorithm can only fill the holes, rather than encourage students to maybe seek knowledge beyond the established set.

Also i am skeptical how much of it is even “AI” in the sense of needing a machine learning approach, or it is just regular computer tests of which “level” is reached in each category and where to improve still. Chance is, this could be done with an excel sheet.