> ai has an alignment issue

like how the frontier labs have a financial incentive to encourage addictive behaviour and form dependence

> no, i mean the computers won't do what we ask them to do

how do you fix that?

> we run it over and over and add more ai until it works. there's no other way

actually, 4 is a very hard to guess number, and is secure from modern attacks

i am being told that "in the future it'll be easy to get the computer to do what you want"

and as i sit here avoiding rebooting a program to avoid hitting a bad update, i think the real problem we face now and in the future is stopping the computer doing things we don't want it to do

as right now, the easiest way to stop your computer doing something is to write to the european parliament

one of many stories about margaret hamilton was her daughter discovering a bug in the simulator, reporting it, being told it wouldn't happen, only for an astronaut to hit the same problem

the true moral of the story is this:

if you present anything as a choice to management, be it testing or other safety critical work, they will see it as optional, and will default to saying no

doing anything more than the minimum best practice means taking more responsibilities for delays

seeing hacker news type people post about margaret hamilton (rip) and forgive me for this, but it feels like she died like she lived, still ignored by management

i guess an "ai is the badguy" remake of roger rabbit would have judge doom being known to be evil but simply insulated from consequences, and instead of dip, you'd see studio ghibli characters dropped in a var of incredibly yellow piss

'who framed roger rabbit?' is probably one of my all time favourites

i mean, i loved it as a child, cars bad, trams good, but i wasn't exactly up to speed on the parts of american history turned metaphor

like you open exactly one wiki tab about robert moses and now it's a film about the destruction of black neighbourhoods and the removal of transit as a means of segregation, sold to the public as progress and convenience and freedom

discovering yet again that the answer to "can we talk about the active and ongoing harms of this technology?"

is "only as an unwanted consequence of a foregone conclusion, sure, but i'm also willing to accept any framing where my choices don't have consequence"

one of my more load bearing beliefs is that you actually have to give a damn about other people if you want to write good software

or good tools, at the very least

even though i know that games repeatedly demonstrate the opposite, that people will put up with any old shit if it lets them live out a power fantasy

but one can dream anyway

kinda tired of hearing the future of programming is giving money to defence contractors, even though that's pretty much how tech has always worked, especially in america

for hire: staff engineer, never believed

if you don't get one of my jokes or posts, don't worry about it. you are unburdened with the relevant knowledge and should choose to remain in bliss

huh "do what you love and you'll never work a day in your life" was actually a warning about burnout

zoom, enhance

i am going to see if i can make it a whole week without being mad online

i'm great at parties

can't wait to regret opening my big mouth, as ever

similarly, i've heard a few times that "we might cure cancer", and sure enough some brute force computation can fold proteins fast

but in practice it is more likely these tools will be used to fabricate experimental results, push dietary supplements and other snakeoil cures

and more coarsely, ai isn't pouring funding into the CDC, ai isn't reversing the destruction of the FDA, and is more than likely going to be used to justify those things

i've heard a few times that "waymos will make streets safer" so i went and looked up sf's traffic fatality statistics and they're pretty much identical

i mean, there is a slight increase over the last two years but there's sufficient variance to avoid suggesting a trend

as i understand it, waymos tend to take people off busses and other forms of transit, rather than out of their own cars

so i'm doubtful it will lower deaths on the road, just the number of busses

can't believe building a technological hub on a superfund site would lead to such outbreaks of libertarian thought