Marquis de Geek
Simone@marquisdegeek I respectfully disagree. New code is nearly never “original“, i.e. innovative and creative. New code gets the job done with existing tools. It is engineering, not poetry. AI is bad at extrapolating, but good at interpolating (it is just ML on steroids). Hence, AI is relatively good at writing SW, as writing code means “interpolate” in the space of functions achievable with existing tools and given instructions. Often AI lacks a good prompt, rather than the potential.
@simone_z Aren't we agreeing - you said "New code is nearly never “original“", as did I?!?
@marquisdegeek well, if you put AI in the reasoning, but you think the same for humans, we can agree that is not really innovation, but still, working code solves a problem, hence it makes sense to write it. But I think you meant to say that generating code with AI should be pointless. If this is the case, we disagree, as I think both human and AI code are useful even if not innovative. Would mind elaborating what you meant?
@simone_z Essentially, why is the same code being written twice?
A library/module could/should be already available, and will include test suites.
@marquisdegeek I am not sure which code you are refereeing to. Do you mean AI is unable to use libraries and modules? AI can use them the same ways software engineers do. AI written code is certainly less creative and sophisticated, but the code it writes is worth to be written, as long as it gets (as often is the case) the job done.