kixiQu
kixiQu on Hacker News
It'll be interesting to see if we get better / worse at figuring out what effective LLM interaction should look like, since that seems prerequisite to this kind of evaluation. Right now there's so little data on whether e.g. casual conversational interaction works better than structured prompting, and so, so, so much superstition - I don't know if I'd really trust judges to have good instincts about what's effective rather than what resembles their own use patterns.
I would like to preemptively acknowledge that this is not the center of the article, but:
> Did you know IQ tests are effectively illegal for hiring in the US because they disproportionately screen out minorities? I learn things when I research these posts.
The provided link discusses how this is itself a common misunderstanding and they're not illegal if they have to do with the job, but you have to justify that they do have to do with the job.
> Using fully automated scoring would also be against the law in some places: New York City requires annual bias audits of automated hiring tools, and you don’t want that heat.
The elision of the gap between "[it's illegal to do this]" and "[it's illegal to do this without checking for bias]" is interesting.
Font-cloning of dubious quality not novel to the LLM era!
https://luc.devroye.org/fonts-27506.html
> Anyone who has a digital "OPTIfont" and a font editor can readily view the problems, including usually several times too many Bezier points within any character. I counted 78 control points on a minimal character, for instance, that should have had less than a dozen.
For a lot of folks, they're never going to notice the difference, but this gets into target audience sophistication, etc. etc.
(edit for clarity: I'm pretty confident the ChatGPT typeface clones would be low quality. Notable too, however, that the Century Library ones might also be!)
Convincing yourself that your local birds aren't sufficiently interesting is a great way to rationalize avoiding getting into birding. (Source: I am trying to learn to distinguish all the pigeons that constitute most of the bird population near me. They're great.)
And it is a classic for folks with limited mobility:
https://www.nytimes.com/wirecutter/blog/birding-window-get-s...
There's a reason why it has a reputation as a great retiree hobby - it's "go sit on a bench and pay attention" not "first, purchase a mountain bike of adequate spec...".
Are you saying that the population that has access to virtual reality hardware is "much larger" than the population that has access to going outside?
This looks like it's capturing how wealth is shifting around from people moving. But how does this net out with changes in wealth / income / whatever from people who stayed put? It'd be very interesting to learn that California was actually contracting, but while this data presentation suggests it, that's not what's actually here....
FTFA:
> Can this be gamed? Teaching the AI to write less like stock Claude and stock ChatGPT? Yes, absolutely. And honestly that would be great. Because half of the irritation is that all these AI-written article sound the same! If they're all being written with each individial's, each company's, own rules and preferences and character that'd probably be nice?
To my knowledge economists do not believe this one is black-and-white, but that's much more of a question about the issuance of H1B visas they begin with anyway.
This is also why I as a US citizen strongly support green cards for everybody in the industry who's fractionally eligible for them - my working conditions are shaped by what they can get the labor force as a whole to put up with, and that's shaped by visa indenture.
That doesn't feel accurate to me. If you're laying off employees en masse while getting new h1b visas, maybe - but everybody I know who's gotten a green card from tech employment was employed and working for a long time before that process even started.
> I don't really see what damage this would do or why you'd have to apologise.
FTFA:
> I didn't learn these libraries. Hell I barely know which libraries I've built on top of! Open source development used to run on attribution, how can I attribute what I barely know I've consumed?! Why would those developers continue to work on these projects when they now go even more unappreciated than they used to be?! I assume many have stopped.
> And those are just the libraries. The code it is generating is synthesized from being trained on code they have sucked out of the public (and perhaps not so public) internet. Code like my own from my own Github page and many many others, all without attribution or compensation. I could have paid someone to build it, freelance software development used to be a thing, but I didn't. I bet that market is pretty dry these days.
Even pre-LLM: I've found a lot of colleagues (mystifyingly to me, both seniors and juniors) who really want to be able to turn their brain off when doing investigations into complicated cross-system behavior (feedback loops abound), and they don't like hearing that it can't be collapsed into a runbook. If a drinking bird (
https://en.wikipedia.org/wiki/Drinking_bird) on an Enter key could do it, I could rig something up so I wouldn't need you!
did you reply three times to my comment?
negative parallelisms everywhere in an advertorial for digital downloads: truly the era of LLM advertising
if this had had the period source material in it, it'd've been fun to see, though. pity.
> If you don't want your content to end up in some database/archive don't publish it for the whole world to see.
This principle somewhat reminds me of the line that "If you're not paying for the product, you are the product", and it seems to me similarly misleading - my data gets harvested and sold by companies with which I have non-paying relationships and by companies I have to pay for things (I am made the product in both cases). As you note - the AI companies are ignoring copyright law and pirating everything that seems useful to them regardless of whether it was published for free access.
The potential externalities here are troubling.
https://vbuckenham.com/blog/how-to-find-things-online/
> Please don't complain about tangential annoyances—e.g. article or website formats, name collisions, or back-button breakage. They're too common to be interesting.
https://news.ycombinator.com/newsguidelines.html
(reader mode works perfectly well on this site afaict)
I really like this kind of thing but I wish there were more tutorials to help get good with keyboard shortcuts. vimtutor was lifesaving
Does this mean it'll be on Bedrock soon? I hear great things about this model but I want AWS data handling practices...
I definitely agree with your sentiment in general. However, I'll note that I find this sans serif to have a nicer feel than many I see these days (even among humanist ones) and it's cool that they've managed to achieve that while being acceptably inoffensive for mainstream use
This seems directed at people who are angry about Things In General. AI as a whole. The Industry. In reality, I think a lot of folks who are the most angry are angry about specifics they observe - tokenmaxxing policies impacting performance reviews, leaders encouraging binning hard-won best practices, coworkers shoving slop in your face, managers seeking funding for crazy AI things everyone knows will fail but look good to VPs while actual critical stuff gets no headcount... It's not looking for "someone to blame" to recognize this stuff around you, and it's by
no means "a comforting story".
Maybe there are lots of people out there who are so vague about what they're angry about that this isn't just massively condescending to the rest of us so much as mistargeted. At the same time I've found that folks who really don't believe in these downsides sometimes don't hear you when you bring specifics.
It's interesting to think how nearly-e-waste level hardware could combine with something [home-cooked](
https://www.robinsloan.com/notes/home-cooked-app/) to be really cool, especially for folks who might have been intimidated by this kind of development before.
I am trying to re-find it but there was a study I'd read that was saying that in general if you don't break it down, there's a massive amount of weight stigma impact on women's employment but not men's. However, if you do break it down, you see that that impact doesn't exist for blue collar men, and does exist (if in a mediated form) for college-educated white collar men. If anyone can help turn this up (or something contradicting it) it seems useful.
I think there's a lot of just-world fallacy going around re: how I keep running across the following pattern online now that GLP-1RA drugs exist.
Commenter A (esp. often a woman, though of any age): I lost a lot of weight and people treat me radically better. I did not change my behavior in a way that explains this; it seems like stigma / prejudice.
Commenter B: I have never met you but it is probably the case that stigma is not relevant and in fact you did somehow become more confident or otherwise start behaving differently in ways that mean your prior and current experiences of the world were somehow deserved.
Which is kind of wild!
This is interesting to me because in my job as a senior dev I can tell that there are greater opportunities for me if I can improve my skills. As a result, I think this may speak to your work environment more than the nature of being senior.
Plausible-sounding mechanisms like you're presenting here do not add value to the discourse without data. There are tons of plausible mechanisms we could make up absent empirical evidence to distinguish them.
My sense is the same as the authors that LLMs + an Anki-like -> mediocrity. But I feel like there
ought to be such potential there! Even just things as simple as rewording the same question a bunch of equivalent ways to avoid recognizing the sentence structure...
Downside for AI potential as a whole, framed broadly: We don't seem to be good enough yet at identifying what friction is functional and what we should strive to automate/eliminate.
I would also support this if there were a way to do it without privacy implications for actual adults. Unfortunately, there is not, and more unfortunately still, there is not widespread understanding of that fact. I think our industry owes more honesty to the public about the trade-offs.