8,254 karma · joined April 30, 2018
It's something people need to be made aware of so they can be prepared for it.
Otherwise you can end up getting sucked into the frame of these haters, stuck in a perpetual loop of defending yourself and raising your blood pressure until you quit, giving up whatever made you interested in the project in the first place.
It's only "falsely" if the courts rule that LLM-written code is not copyrightable, and so far there's no indication of that happening, although I do see a lot of people secretly wishing for that to be true.
They are trained on a narrow set of data and don't really understand the real world.
If you let them run wild with no supervision they'll turn your company into slop.
Someone needs to be there to reign them in, asking the all-important questions like "is suing all of our customers and making them angry REALLY the best option?" (to which they will reply, "You're right to push back on this").
This is typically the role of the CEO, however, as we know, most CEOs of large successful companies are too busy to actually do that. They're mostly training Brazilian Jiu Jitsu, shitposting on their own social media platform, trying to manipulate international politics, or helping run their family nonprofit.
This is why I believe that legal departments need to be demoted. The legal department should not report to the CEO, they should report to a new role that is above both PR and legal. And that person's job should be to force legal and public relations (which we know are natural enemies) to work together. Every press release goes through legal now, and it's only logical that ever legal action should go through public relations review as well.
The fact that public relations at most large companies has atrophied from disuse in our public-equity-owned unicorporate cooperation-over-competition chaebol/zaibatsu/conglomerate world is a problem for another day.
The first is an XKCD 605 (Extrapolating). The author makes bold predictions for the future as though they're inevitable, but gives zero evidence, because there is none, we're still at T=0. IBM is still in business. Wake me up when it's been taken out by someone vibe-designing a mainframe in kicad. Or even, let's see a dropbox clone get off the ground. Should be easy, right? (It's an HN tradition, after all.)
The second mistake is mistaking a large multiplier for infinity. The cost of code has not "collapsed to almost nothing" as the author claims. Talk to your CTO/CFO if you want the real story. They're probably freaking out over AI spend. But the point is, yes, AI has made things cheaper, but they didn't become monotonically zero.
AI had also raised the bar. You can't ship "good enough" anymore, it has to be great. And great still takes time. Time in QA, time obsessing over every workflow, time talking with customers and planning features.
Taste hasn't even entered the picture, and the analysis is already wrong.
That's an interesting historical fact, and one that we've locked ourselves out of. I looked into stripping Linux down to the bare minimum, and it's extremely difficult to untangle the codebase. There are inter-module dependencies that aren't specified anywhere, the build just mysteriously fails. The kernel config tool pretends to give you a lot of options, but in reality a lot of configurations will never actually compile, and a lot of things that appear to be removable actually leave huge chunks of code behind because they're used by other things.
And that's just the kernel. If you want X server, browser, text editor, you'll be surprised at what dependencies they require. If I recall correctly, Xorg requires MESA which requires... drumroll please... clang and llvm (about 1GB of code if you clone from github). It's madness.
Do what you have to do. shrug But when it doesn't have any effect, consider that maybe your bar for "bad faith argument" includes large swaths of completely legitimate differing opinions.
> log every trip
> This system tracks the whole of everyone's movements, every day
Funny, because current lawsuits over flock have hinged on the point that they DO NOT have total coverage of roads. Author seems to be talking about Long Island, which is an unusual case, and flock might have total coverage there. But special places need special laws, and an edge case isn't the model for the whole world!
> The Supreme Court held in Carpenter v. United States that the government may not track the whole of a person's movements without a warrant.
No, that court case established that the government needs a warrant to get, specifically, historical cellphone pings from a 3rd party (the cellphone companies).
And overly-emotional language to fool the more credulous:
> It is called Flock for a reason. We are sheep only if we let this happen
Oh please, it's named flock because it's perched on a pole, like a bird, not a sheep.
> feed a national database searched by 59 outside agencies nobody in this town chose
> 0: Opt-in or opt-out decisions the Town made about any of this sharing, in its own words
Well why not? It seems like a lot of the flock discussion takes a bizarre twist where people go "It's useless to try to limit it, we need to ban it outright". Like okay? Do we ban the police when one goes rogue? (I guess that's kind of a bad question to ask, given that "defund the police" was trending a few years ago)