> in practice, often it doesn’t happen, people make assumptions that the code works
True, but that's an inexcusable practice and always has been. We as an industry need to stop accepting it.
True, but that's an inexcusable practice and always has been. We as an industry need to stop accepting it.
All of us rely on millions of lines of code that we have not personally audited every single day. Have you audited every framework you use? Your kernel? Drivers? Your compiler? Your CPU microcode? Your bootrom? The firmware in every gizmo you own?
If "Reflections on Trusting Trust" has taught us anything, it's turtles all the way down. At some point, you have to either trust something, or abandon all hope and trust nothing.
Of course not. I exclude the CPU microcode, bootrom, and the like from the discussion because that's not part of the product being shipped.
But it's also true that I don't do a deep dive analyzing every library I use, etc. I'm not saying that we should have to.
What I'm saying is that when a bug pops up, that's on us as developers even when the bug is in a library, the compiler, etc. A lot of developers seem to think that just because the bug was in code they didn't personally write, that means that their hands are clean.
That's just not a viable stance to take. The bug should have been caught in testing, after all.
If your car breaks down because of a design failure in a component the auto manufacturer bought from another supplier, you'll still (rightfully) hold the auto manufacturer responsible.
That’s reacting to a bug you know about. Do you mean to talk about how developers aren’t good enough at reacting to bugs found in third party libraries, or how they should do more prevention?
In this case, it seems like OpenAI reacted fairly appropriately, though perhaps they could have caught it sooner since people reported it privately.
“Holding someone responsible” is somewhat ambiguous about what you expect. It seems reasonable that a car manufacturer should be prepared to do a recall and to pay damages without saying that they should be perfect and recalls should never happen.
My point was neither of these. My point is very simple: the developers of a product are responsible for how that product behaves.
I'm not saying developers have to be perfect, I'm just saying that there appears to be a tendency, when something goes wrong because of external code, to deflect blame and responsibility away from them and onto the external code.
I think this is an unseemly thing. If I ship a product and it malfunctions, that's on me. The customer will rightly blame me, and it's up to me to fix the problem.
Whether the bug was in code I wrote or in a library I used isn't relevant to that point.