This is like arguing that the Model T demonstrates that cars will never be safe enough to drive.

Yes, there are ways to do this wrong. We have a much better idea what they are and our standards for precision and reliability are much higher. The 1940s are long gone and we aren't the baby boomers. We require seat belts in all cars, houses to be built to fire and electrical codes, and household chemicals to have plain labels and clear warnings.

We also now know that we need to specifically target something like three sub-species of mosquito and that if we hit exactly those three the ecological impact will be essentially zero; we've verified that they don't compete with anything and that nothing eats them and that they don't interact in any of the dozens of ecological systems that professional biologists understand that programmers never learn about. We know that we need to hit exactly those sub-species because DDT happened. We know that we need to study things like evolution rates and virulence before deploying engineered organisms; we have, after all, all watched Jurassic Park.

The people working on this understand what will happen if their solution isn't good enough. You are not adding anything to this situation except uncharitable negativity. You cite DDT. I'll cite Greenpeace, whose blind opposition to nuclear power is still contributing to the ongoing death of our planet. Don't be Greenpeace.

Your counter examples are about the mitigation of unexpected and unwanted effects. These were implemented after the fact.

There are situations that have no possibility of redress.

Your paragraph about ecological impact can serve as an example of the lurking danger. I wasn't even concerned about ecological impact. My concern was about artificially creating a resistant species. That's one of the pitfalls here. We're talking about the same thing, but because it's new we're not addressing the same issue.

We got away with DDT because:

- We could stop using it.

- There were alternatives that reshuffled the genepool of a now largely resistant population the other way, after the fact.

These two are not a given when deploying engineered organisms. Even though both are means to fight malaria, they are not the same and need different considerations of which some are new and not valid for any older approaches.

> The people working on this understand what will happen if their solution isn't good enough.

So do most programmers...

> I wasn't even concerned about ecological impact.

I spat out a post in ten minutes that looked further out than you did. I wasn't even trying.

The people that actually do this stuff are trained, credentialed biologists and ecologists that have taken courses on this and spend literal decades reading studies about proposals and going down checklists of things that have caused problems in the past. It is literally their job to make sure nobody can say "You didn't consider X" during review, on the news, or in court.

Back in tenth grade I took an honors biology course. The year's final project was to write an environmental impact statement for, if I recall correctly, chopping down twenty or thirty trees in a nearby park. In retrospect, we worked off what I suspect was a massively stripped-down version of a tiny fraction of the real rules.

It took five of us a month and we wrote a hundred pages single-spaced.

> So do most programmers...

I certainly won't argue that the field of software engineering doesn't need more rigor - there are reasons that I prefer writing cleaner code than my coworkers often do. But this proposal isn't operating in the same regime I usually do.

I mangle data at scale for a fuzzy rules-engine type thing. I bet you do something with comparable correctness requirements. There are times I've looked at a bug, looked at the logs, figured out the underlying conditions, estimated the future occurrence, concluded that I spent more money just looking at the logs than it would cost to remediate the data by hand for the next ten years, and estimated that it'd take me three times as long to actually fix it. I do know what happens when my solutions aren't good enough, and I bet you do too.

Aerospace and defense programmers operate in a regime far more like what we see with these environmental proposals. And guess what? They know what happens when their code isn't good enough... and they have huge checklists that they go down for every single change and have their code audited regularly and tested obscenely thoroughly and spend weeks tracking down tiny bugs, stamping them out, and adding them to the checklist.

(Before you bring up the recent boeing failures: Take a look at the politics and incentives. Where did the economic and political incentives lie there? Where do you think the economic and political incentives lie for a proposal like this? Especially given how much environmental fearmongering we see as a result of movies like Jurassic Park? Do you think that anyone would dare approve a project like this with anything less than every single i dotted and every single t crossed?)

> but because it's new we're not addressing the same issue.

Development of resistance is such a well-known failure mode for pest-control tools that wikipedia's page on DDT has a whole second-level paragraph titled "Mosquito resistance".

Why do you think that evolution and resistance are novel failure modes, and why do you think they'd have been missed in the analysis?

> that looked further out than you did.

Very much so. You were presenting how there would be no danger to the ecosystem if these three sub-species were successfully destroyed. I never got as far because I wondered about the similarities of population regrowth after destruction with insecticide and destruction with modified organisms.

> Before you bring up the recent boeing failures

I don't need to. Building nuclear facilities is (usually) a very rigorous field with the highest quality of engineering, but sometimes factors of design considerations interact in ways that create new conditions that were impossible to predict. And since you mentioned aeroplanes: if you look at NTSB reports, you often find the cause of crashes is a long chain of trivial events, every single one of them meticulously checklisted and tested with fallback modes to avert disaster. It's not the components, but the intractability of the web of interaction.

> Why do you think that evolution and resistance are novel failure modes

Because evolution and resistance are novel failure modes in the context of genetically designed natural enemies. Previous statistical models might not hold even though the situations modeled appears exactly similar.

> why do you think they'd have been missed in the analysis?

Genetic modification is a statistical process, much like an AI network where something goes in and the output is measured for fitness and fed back. Random artifacts are to be expected and accounted for, but cannot be known to exist before the fact. As such their model might be complete, but it might not be sound.

As you implied, rigor is important, and faced with potentially irreversible consequences, skepticism is not an unnatural or irrational response.