Thanks. In a broader sense, with "AI as code" I mean any situation where we ask an AI model for answers or decisions where we otherwise might have written a program to solve it. See also "LLM functionalism" in the…
Author here. Serious systems undefined, fine - I'd view them as systems that take decisions, not just make predictions (see final quoted line!). "Want to trust a system at the same level as ... a theorem" - straw man,…
Author here. Fair enough on my industry experience. But I hope components, unit testing, regression testing, etc aren't as easily dismissed in real SE environments - no trouble believing formal methods and verification…
Author here. Unfair point, article makes clear I don't expect "outdated verification methodology" to catch up - and I even link that pessimistic expectation with the issue of emergence that fatally undermines NNs. Point…
Author here. Disagree on not slowing down. Indeed "dead end" isn't really the focus of the argument, editor's choice of title, not mine. Focus of argument is these systems are foundationally unreliable so can't be used…
Author here. Fair enough. The "dead end" in the title isn't mine, extracted by editor from what's essentially a side comment on the limits of LLM scaling. My title would have been something like "We can't use current AI…
Author here. Hence the insertion of the word "systems".
Author here. Putting a lot of words in my mouth there. In particular, I don't talk about whether AI is useful for software development - I talk about whether AI is useful as reliable software. I don't discuss how AI's…
Author here. So what! I am not talking about promising directions for AGI, I am talking about having computer systems that we can have confidence in. Sure, AGI if it ever happens will look more like emergence than…
Author here. Nope, that is not my concern about humans in the loop. My concern on that is that any human in the loop has to reconstruct by themselves from the inputs whether the output makes sense, the system provides…
Thanks. In a broader sense, with "AI as code" I mean any situation where we ask an AI model for answers or decisions where we otherwise might have written a program to solve it. See also "LLM functionalism" in the…
Author here. Serious systems undefined, fine - I'd view them as systems that take decisions, not just make predictions (see final quoted line!). "Want to trust a system at the same level as ... a theorem" - straw man,…
Author here. Fair enough on my industry experience. But I hope components, unit testing, regression testing, etc aren't as easily dismissed in real SE environments - no trouble believing formal methods and verification…
Author here. Unfair point, article makes clear I don't expect "outdated verification methodology" to catch up - and I even link that pessimistic expectation with the issue of emergence that fatally undermines NNs. Point…
Author here. Disagree on not slowing down. Indeed "dead end" isn't really the focus of the argument, editor's choice of title, not mine. Focus of argument is these systems are foundationally unreliable so can't be used…
Author here. Fair enough. The "dead end" in the title isn't mine, extracted by editor from what's essentially a side comment on the limits of LLM scaling. My title would have been something like "We can't use current AI…
Author here. Hence the insertion of the word "systems".
Author here. Putting a lot of words in my mouth there. In particular, I don't talk about whether AI is useful for software development - I talk about whether AI is useful as reliable software. I don't discuss how AI's…
Author here. So what! I am not talking about promising directions for AGI, I am talking about having computer systems that we can have confidence in. Sure, AGI if it ever happens will look more like emergence than…
Author here. Nope, that is not my concern about humans in the loop. My concern on that is that any human in the loop has to reconstruct by themselves from the inputs whether the output makes sense, the system provides…