We Need an FDA for Algorithms
nautil.us
nautil.us
I can't say for sure that the author or interviewee even knows what the definition of an algorithm is.
No, the government absolutely should not be involved in regulating algorithms. You'll be able to enjoy a burdensome system for anyone that isn't a massive corporation. Good luck competing against Google when every 10 lines of code you write, you need to seek government approval.
I know you're not supposed to be dismissive on Hacker News, but it's so difficult to take this seriously when you display this level of technical ignorance.
A solution is to break up and/or regulate FAANG sufficiently such that algorithmic regulations do not impede David’s from reasonably competing.
Or, such regulations could also be targeted at bigger players past a certain size, and not apply as strictly to smaller scale applications. It is the broad scale application of an algorithm that often facilitates its ability to be effective, after all.
The phrase "unchecked algorithms at scale" does not mean anything, it's just written to look scary.
Now the discussion of breaking up (or regulating) tech companies appears to have expanded to Apple and Netflix (re: FAANG) for some reason.
AFAIK, no one else outside of people working at Facebook can see how the algorithm decides what should be on top of your feed.
I feel confident that the author has also not gone through an FDA approval process. My wife works in regulatory affairs for a medium sized pharma. There are a lot of hoops to jump through, egos to stroke, and huge costs involved. Want to blow a few weeks and a couple $million because a picture in an ad might be showing someone smiling too much (you can't imply a drug will make you too happy unless is indicated for such), go to the FDA. Don't get me wrong, the FDA exists for a reason, and is far, far better than not having oversight.
Even if (beyond all logic) a reasonable case could be made for government oversight, the FDA is the wrong model.
"Hannah Fry points out something interesting about the phrase “Hello World.” It’s never been quite clear, she says, whether the phrase—which is frequently the entire output of a student’s first computer program—is supposed to be attributed to the program, awakening for the first time, or to the programmer, announcing their triumphant first creation."
Did she try asking programmers instead of guessing?
Jest aside, I think there is a real lack of clarity in what Fry is asking for here. You don't need a computer to bumble your way into reinventing redlining, yet at the same time a single algorithmic tool can be benign or morally terrifying based on application.
What are the boundaries Fry seeks and what moral principles do we base them on? What mechanism(s) does she envision as proper to enforce them? That bit about pre-market approval isn't entirely a joke - the analogy to FDA specifically rather than another regulator implies either pre-market approval processes or a bad author. How do we categorize which pieces of software to subject to what levels of scrutiny?
I don't mean to poo-poo the idea of getting limits around software that is making more and more impactful decisions. Hopefully the reason this article is so light on detail is because the reporter doesn't understand the subject matter. But Fry's response to the data ownership question is the sign of someone that always punts to "experts" when pressed on the finer details.
And of how many government deployed algorithms should be documented and evaluated, as a condition of use. For example, "risk" profilers that are being used to determine prison sentencing and parole/supervision provisions.
Health care related algorithms are another obvious candidate.
Question is: why then are they calling them algorithms.