Hacker News Reader

Comment by SketchySeaBeast | Hacker News Reader

The most chilling statement out of the article to me is "Musk’s impatience with Neuralink has grown as the company, which launched in 2016, has missed his deadlines on several occasions to win regulatory approval to start clinical trials in humans, according to company documents and interviews with eight current and former employees."

To think a company that is apparently being forced to cut corners and move faster, resorting to "hack jobs" on animals in order to show progress, wants to test on people turns my stomach. How do you effectively shift from that negligent approach to the careful diligence demanded when testing with people? What happens when the human trials take too long to show progress?

View on HN

> How do you effectively shift from that negligent approach to the careful diligence demanded when testing with people?

Presumably as easily as people turn from casually eating beef to raising their children without suffocating them at the slightest inconvenience? People have very different standards of care for humans and animals.

These animals are intended to simulate humans for purposes of research, though. A disregard for doing the experimentation properly has the potential to carry over into production usage, just like coding newbies who write code with SQL injection holes saying "eh I'll fix it later" are practicing writing insecure code.

"Oops wrong glue/vertebra/implant" folks are not people I want anywhere near my brain surgery.

But they are proving they can't even be bothered to meet the much lower standard of care required for animals.

Presumably they are utility maximizers who assume that the 'standards of care' for animals won't actually be rigorously enforced.

Humans sue.

So do the parents of children but it hasn't stopped flaws in Tesla's FSD. Not entirely directly comparable, sure, but it's the same attitude of "get it to market before it's ready".

"FSD", surely

If there's one thing we've learned from Elon's acquisition of Twitter, it's that his default response to "you'll get sued" is a confident belief he can beat the rap.

Precisely. Sounds like an episode of Black Mirror.

Turns out, Agile doesn't work in healthcare.

It can work rather poorly in software from a quality perspective too, but in software the game is often to test a business hypothesis very quickly or ship features fast enough to win feature bingo with competitors and quality often goes on the back burner.

It's just that in health care the bar for MVP is way way higher.

I have a whole rant about how consumer software and software for serious things (ie: people can die) need to be treated differently just like civil engineering and building a shed in your back yard should be treated differently. There is no room for agile code in a finished product where lives are on the line. Its fine for non life or death things and its a great way to prototype ideas. But the control system on your car should be developed with more rigor than that.

As it turns out, "Move fast and break software" is in the vast majority of cases less unethical than "Move fast and break people".

It actually does, but in a measured way. You have to have a hybrid approach. Agile for stuff like little software features and bug fixes, waterfall for big important (often safety and effectively tied) milestones they often precede an animal or human study. Iterate the smaller less safety critical features (which are often software based) and then align the exit of one of those phases with a hardware freeze that captures key safety and effectivity in the hardware and then do a cadaver, animal or human study to prove it. You don't want to apply agile to the hardware usually.

Move fast and break people.

Moving fast and break things actually seems like a decent idea in healthcare BEFORE you get to the human trials, then you should slow down. Not sure how hard a change in culture that will be.

We kill hundreds of millions of animals a year after incredible suffering for much worse reasons than making the blind see. And healthcare moves very slowly atm, leading to a lot of suffering.

I agree with the idea in general but some of the errors are just bad science that doesn’t discover anything. Putting the wrong sized chip in a brain? Using a mistaken vertebrae? Not euthanizing as soon as it’s apparent the trial is too flawed to continue? These are just sloppy mistakes.

Yeah, depends whether the checklists slow things down more than the errors.

Reply on news.ycombinator.com