Part of it is how we scope law enforcement's job. Thanks to technology there's tremendous growth in the amount of information about people. Traditionally law enforcement needs to find evidence that a person may have committed a crime, not necessarily prove a crime. Limited availability of data makes it less likely that law enforcement finds evidence on innocent people.
My take is that even with a warrant, law enforcement's reach into private data should be limited for a free society.
One other distinction to make is that auto-update typically runs the same code on a lot of computers, rather than a RAT which is good at running code on a specific computer.
There's still room for problems... the auto-update could deliver a special version for special people or deliver a version that has special code targeted at run-time, but it's not as easy. And I'd love to see work on minimizing problematic updates as well.
There's a lot of good advice here for dealing with constructive code reviews. They provide valuable learning opportunities.
Also be aware that not all code reviews are good. Engineers have opinions and may wield code reviews as a tool to impose theirs on others. In those cases it doesn't reflect you or the quality of your work.
- We don't really care if we further damage the hair cells that don't work.
- Outer hair cells amplify sounds. If they don't work, well, gotta amplify them some other way.
- A phenomenon called "recruitment" can cause loud sounds to be more painful for those with hearing loss. So you may not be able to amplify everything.
Hearing aids are a bit more sophisticated than an amplifier; they also act as an equalizer, compressor, feedback cancellation, and a bunch of other features of dubious benefit.
"You would expect that if we were getting it wrong, we would be getting it wrong the other way, delivering early."
How likely is it that a surprise works in your favor? In my experience what you don't know rarely does. So my mental model is that completion times aren't normally distributed; they have a long tail. Additionally a task can't take negative time. This makes estimation tricky because a task can take much longer than the average. At my company we typically report an average and then explain why things took longer.
It's hypothetical, but other code may expect a bigger buffer than was actually allocated.
Suppose we're on a 32 bit platform and num is 0xF0000001. When multiplied by 64 (0x40), we'll end up allocating a 64 (0x40) byte buffer. But other code converting num to unsigned may be expecting a buffer large enough for 0xF0000001 64 byte records. After all, that was probably the reason for multiplying by 64.
- If someone else bought an audio book, I would like to be able to access it. I can't if it's audio-only.
- If I want to buy and listen to an audio book, I wouldn't be able to understand the speech without seeing the text. The text gives my brain additional information to process speech. This is much more relevant for music where lyrics help me hear the vocals in the music.
At least according to the article it was a keyword ban, not keyword suggestions. To my reading the intent of the ban is to (1) avoid the appearance of carrying advertisements of sex with children and (2) discourage such advertisements.
The critical view is that this is intended to help child traffickers write clandestine ads so they can continue using the platform however that seems unlikely. The problem with any keyword-based filter is false-positives such as kink content. Such filters operating opaquely and dropping bans in other contexts such as banking and copyright draw deserved criticism. We should be no less critical here.
In the past HA companies had a problem: audiologists weren't technical and didn't know how to program the increasingly sophisticated hearing aids. So the fitting software (1) automatically produces a reasonable fit and (2) at least the ones I've used have a section that asks questions and makes adjustments based on the answers.
It's also worth noting that there isn't really a "correct prescription" but rather it's an art of arriving at settings based on the client's goals and feedback. The audiogram is a diagnostic tool but only a starting point for determining HA settings. From the HA dispenser's perspective, adjusting settings costs them time and therefore money after they've made the sale. From the client's perspective, you'd need to visit the audiologist for every adjustment. Self-programming avoids the problem.
The programming hardware, software, and used hearing aids are all available on ebay. The prices are such that your average HN reader wouldn't have to worry too much about replacement costs if a HA breaks or goes missing. And the flexibility is great too.
You may not need NOAH. For a Phonak hearing aid you only need the iCube II (programming device) and Phonak Target (programming software). The NOAH software is an optional place to store customer data which is useful in a clinic that sells hearing aids from different manufacturers.
I have Phonak HAs so I can buy the programmer (iCube) and software (Target) off ebay for $450 and a pair of hearing aids two generations old for $500-$1000. And if they break or I lose them, buy a replacement and the total is still less than buying from an audiologist.
It's also more flexible. Hearing feedback? Run the feedback test/modeling program. Can swap hearing aids between ears. Friend lose a hearing aid? Let them borrow yours (suitably programmed) and use foam inserts for earmolds (can buy or take a soldering iron to foam earplugs).
As a deaf person, hailing a ride by phone is much better than a taxi. But I also have friends who use wheelchairs frustrated by Uber's unresponsiveness to addressing accessibility; Uber just claims the drivers are independent contractors and so it's not Uber's responsibility.
I felt Doctorow's piece addressed the nuance of Uber bringing both improvements and regressions.
My take is that even with a warrant, law enforcement's reach into private data should be limited for a free society.