Need to let loose a primal scream without collecting footnotes first? Have a sneer percolating in your system but not enough time/energy to make a whole post about it? Go forth and be mid: Welcome to the Stubsack, your first port of call for learning fresh Awful youāll near-instantly regret.
Any awful.systems sub may be subsneered in this subthread, techtakes or no.
If your sneer seems higher quality than you thought, feel free to cutānāpaste it into its own post ā thereās no quota for posting and the bar really isnāt that high.
The post Xitter web has spawned soo many āesotericā right wing freaks, but thereās no appropriate sneer-space for them. Iām talking redscare-ish, reality challenged āculture criticsā who write about everything but understand nothing. Iām talking about reply-guys who make the same 6 tweets about the same 3 subjects. Theyāre inescapable at this point, yet I donāt see them mocked (as much as they should be)
Like, there was one dude a while back who insisted that women couldnāt be surgeons because they didnāt believe in the moon or in stars? I think each and every one of these guys is uniquely fucked up and if I canāt escape them, I would love to sneer at them.
(Semi-obligatory thanks to @dgerard for starting this.)
I had to use clipchamp for something recently and my god, what an awful, enshittified piece of software. Itās sending me emails now!
tangentially: Iāve been getting reminded of a bunch of services existing, by way of pointless āyour year in reviewā bullshit
fuck spotify for starting that misfeature, and fuck everyone else for falling over themselves to get On Trend
And, whilst Iām here, a post from someone who tried using copilot to help with software dev for a year.
I think my favourite bit was
Donāt use LLMs for autocomplete, use them for dialogues about the code.
Tried that. Itās worse than a rubber duck, which at least knows to stay silent when it doesnāt know what itās talking about.
https://infosec.exchange/@david_chisnall/113690087142854474
(and also https://en.m.wikipedia.org/wiki/Rubber_duck_debugging for those who havenāt come across it)
deleted by creator
Interesting article about netflix. I hadnāt really thought about the scale of their shitty forgettable movie generation, but there are apparently hundreds and hundreds of these things with big names attached and no-one watches them and no-one has heard of them and apparently Netflix doesnāt care about this because they can pitch magic numbers to their shareholders and everyone is happy.
āWhat are these movies?ā the Hollywood producer asked me. āAre they successful movies? Are they not? They have famous people in them. They get put out by major studios. And yet because we donāt have any reliable numbers from the streamers, we actually donāt know how many people have watched them. So what are they? If no one knows about them, if no one saw them, are they just something that people who are in them can talk about in meetings to get other jobs? Are we all just trying to keep the ball rolling so weāre just getting paid and having jobs, but no oneās really watching any of this stuff? When does the bubble burst? No one has any fucking clue.ā
What a colossal waste of money, brains, time and talent. I can see who the market for stuff like sora is, now.
ai fan asks chempros about their use of lying boxes: majority opinion is that this shit is useless, leaks confidential information and is a massive legal liability https://www.reddit.com/r/Chempros/comments/1hgxvsj/ai_in_the_workplace_how_have_chemistsscientists/
top response:
Itās a good trick to be instantly dismissed. No, really, thatās the latest I had in terms of company policy. If youāre caught using AI for anything, youāre out the door. Itās a lawsuit waiting to happen (and a lawsuit we cannot defend against). Gross misconduct, not eligible for rehire, and all that. Same as intentionally misrepresenting data (because it is). (Pharma)
Days since last comparison of Chat-GPT to shitty university student: zero
More broadly I think it makes more sense to view LLMs as an advanced rubber ducking tool - like a broadly knowledgeable undergrad you can bounce ideas off to help refine your thinking, but whom you should always fact check because they can often be confidently wrong.
Seriously why does everyone like this analogy?
good question, i have no clue especially that i wasnāt like this as undergrad, itās really not hard to say āi donāt know, bossā or āmore experimental data is neededā and chatgpt will never say this
shitty undergrad wonāt probably leak confidential info either (maybe on sender side, but never on receiver side, as in receiving unexplained stolen confidential info from cosmic noise)
From the replies:
In cGMP and cGLP you have to be able to document EVERYTHING. If someone, somewhere messes up the company and authorities theoretically should be able to trace it back to that incident. Generative AI is more-or-less a black box by comparison; plus how often itās confidently incorrect is well known and well documented. To use it in a pharmaceutical industry would be teetering on gross negligence and asking for trouble.
Also suppose that you use it in such a way that it helps your company profit immensely andāuh oh! The data it used was the patented IP of a competitor! How would your company legally defend itself? Normally it would use the documentation trail to prove that they were not infringing on the other companyās IP, but you donāt have that here. What if someone gets hurt? Do you really want to make the case that you just gave Chatgpt a list of results and it gave a recommended dosage for your drug? Probably not. When validating SOPs are they going to include listening to Chatgpt in it? If you do, then you need to make sure that OpenAI has their program to the same documentation standards and certifications that you have, and I donāt think they want to tangle with the FDA at the moment.
Thereās just so, SO many things that can go wrong using AI casually in a GMP environment that end with your company getting sued and humiliated.
And a good sneer:
With a few years and a couple billion dollars of investment, itāll be unreliable much faster.
for anyone wondering cgmp/cglp means current good manufacturing/laboratory practices and itās mostly a set of paperwork concerning audits etc and repeatability of everything
AI could be a viable test for bullshit jobs as described by Graeber. If the disinfotmatron can effectively do your job then doing it well clearly doesnāt matter to anyone.
idk, genai can fuck up a couple of these too
Itās not an exhaustive search technique, but it may be an effective heuristic if anyone is planning The Revolutionā¢.
In further bluesky news, the team have a bit of an elon moment and forget how public they made everything.
https://bsky.app/profile/miriambo.bsky.social/post/3ldq2c7lu6c25 (only readable if you are logged in to bluesky)
the team have a bit of an elon moment
āOh shit, which one of them endorsed the German neo-Nazis?ā
Aaron likes a porn post
āWhew.ā
Not A Sneer But: āPrinc-wiki-a Mathematica: Wikipedia Editing and Mathematicsā and a related blog post. Maybe of interest to those amongst us whomst like to complain.
very interesting, thank you for sharing
Yāall, with Proton enshittifying (scribe and wallet nonsense), I think I am never going to sign up for another all-in-one service like this. Now I gotta determine what to do about:
- Proton Mail
- Proton VPN
- Proton Drive
- Proton Calendar
and Iād be forced to reassess my password manager if hadnāt already been using BitWarden when Proton Pass came out.
Self-hosting is a non-starter (too lazy to remember a new password for my luggage). Any thoughts? Are other Proton users here jumping ship? Should I just resign myself to using Proton until they eventually force some stupid ass āChatbot will look at the contents of your Drive and tell you which authorities to surrender yourself toā?
I am no tech expert but I use tuta for email and disroot for forms, pads and file sharing.
For VPNs, at least, I can offer some suggestions. If you wanted to securely access a specific box or network of yours, tailscale is pretty great and very painless to use. If you wanted to do stuff without various folk noticing then thatās a bit trickier but Iāve been happy using mullvadā¦ theyāre not the cheapest, though they have some splendid anonymous payment mechanisms (you can literally mail them a wad of banknotes with a magic code on a bit of paperā¦ you donāt even need to muck about with bitcoin).
I have a subscription for Private Internet Access that I was using before subscribing to Proton Mail (which comes with Proton VPN). I figured it was all the same (they all have a slightly skeezy feel to me).
Then I checked out Mullvadās website and itās really quite awesome. Everything about their service has a āwe want to make this accessible to everyoneā vibe, which I appreciate. I am going to try it out. <3
Oh yeah I forgot to mention that in my comment: drop PIA. Never touch anything owned by PIA or Kape. Ever.
Yeah I stopped using it after switching to Proton.
also, how are you liking bitwarden?
I really need to kill off my current password manager and bitwardenās looking like the least worst of current options (esp. when paired with something like vaultwarden instead of running a fucking nodejs sync server on the internet), but also some of it seems quite stunted[0]
itās gotten so bad that Iāve started pondering writing my own, because good god does basically every option out there depress me
[0] - no global hotkeys? the fuck
They have a CLI app though which you can hook up to dmenu or rofi or whatever to get global shortcuts.
https://github.com/firecat53/bitwarden-menu
Their desktop app is a bit shit anyway. I just use the CLI and the Firefox extension and itās working solid.
alas: my main workstation is (non-slate) macos, and itās unchangeable for the foreseeable future
good to know those (already) exist as options, though. if I can find some spoons Iāll try look around and see if thereās maybe something similar I can hack up/agglutinate from whatās around
Their desktop app is a bit shit anyway
I havenāt even tried it yet because Iām real āehhhhhhhhhhhhhhā about even the idea of a js-/ts-based gui client for my password manager. largely because Iāve met too many js/ts devs and I outright donāt trust their competence and processes. so your post is definite motivation for me to eyeball some of the other clients too
also, how are you liking bitwarden?
I am happy with it. That they only charge $10 a year for services I donāt even need (I could use a separate 2FA app) and allow you to self-host is a good sign. I plan to eventually set up a workflow in Sway (Wayland tiling WM) with a CLI tool (e.g. https://crates.io/crates/rbw, or the official one), so the interface is not terribly important to me. I would definitely recommend trying a free account to see if it fits into your workflow.
itās gotten so bad that Iāve started pondering writing my own, because good god does basically every option out there depress me
I am in the same boat, except all of the software Iāve ever written has been TeX, or giving contrived examples to undergrads to demonstrate why
dp[i][j]is a shit table name or whyāis better thanfloat('inf')orMAX_INTin pseudocode. So I am only theoretically up to the task, which is ā¦ IDK maybe I should start grifting?But for real, I have considered writing my own:
- VPN client where we donāt have to jump through the hoops of learning a new shitty client, or finding out that their client runs like ass in Linux (Proton)
- Password Manager
- Config editor, so I donāt have to edit
/home/${USERNAME}/.config/sway/config.d/90-fuckyou-this-is-where-we-keep-system-suspend-shit.confevery time I want to change something. āOh no you gotta edit the Kanshi config for that one.ā Itās tedious to remember where various programs look for the config and whatever particular syntax is chosen (isnāt this fucking solved withtomlfiles already?) - An Android reminder app that isnāt some stupid Taylorist metric-worshipping bullshit.
PS: There is Goldwarden which I know absolutely nothing about but looks neat. It does suggest that you could just write your own that is bitwarden compatible.
I am in the same boat, except all of the software Iāve ever written has been TeX
Iām sorry
giving contrived examples to undergrads to demonstrate why
dp[i][j]is a shit table name or whyāis better thanfloat('inf')orMAX_INTin pseudocodethat sound you can hear is my despairing screaming[0]
VPN client where ā¦ jump through the hoops of learning a new shitty client
(not a pitch, but multiple commercial references) I really liked how simple tunnelbear made this for a lot, and also quite like how slick the wireguard desktop-style handling is (you can see this for example with fly.ioās integration to that). I think thereās long context here, and if you buy me a beer I could rant in detail
PS: There is Goldwarden
oh good, itās in Go, my other code allergy
shitposting aside, re the password manager thing: @self and I have co-ranted in dms, and about similar gripes.
so, by way of idea, loose laundry list for foundations/design: modern crypto (jfc why is so much still going āyeah gpg is fineā), crdt sync, a sane fucking language to build everything on, own-devices friendly (in the āyou can sync device to device peer-wiseā sense, vs the āthereās a remote server brokerā sense), and pretty okayā¢ interfaces for client building/extensibility
Iām sorry
me too, also i lied/forgot to mention that my particular PhD situation is so fucked up that i went from pure mathematics to cuda
GPUs: not even once
That is a good rule. The GPU programmers seem to think this is good code and that itās well-documented. I am still pretty out of my depth in this field, but it feels so silly to me. There is this historical bullshit about fortran only allowing 5 characters for a function name, and that (combined with some appeal to domain-specific knowledge) is used to justify stupid, freshman level shit like
if uplo == 'U': # manually fill in this part with the version of the algorithm that is for upper triangular matrices else: # just assume it's always U or L without checking, god forbid you use something modern like an enum, or even just a boolean # manually fill in this part with the version of the algorithm that is for lower triangular matricesedit: if memory serves, booleans were first discovered in 2011 by John T. Boole, which is why they donāt show up in fortran
last time it came up, tuta was the least worst of the mail options. itās not the same offering as protonās in-garden encrypted, but nothing is afaik. rest of it is pretty okay (I have some (not all[0]) domains on there)
the rest of the things I donāt have a direct recommendation in part because [0] and in part because I donāt use computers entirely like how a lot of people do. that said
storage: backblaze storage pricing is not bad. they might have a desktop app thing? calendar: caldav is a dark art beyond my ken - I havenāt even got that shit playing nice on my own things[3]. fuck knows who does this well. vpn: mullvad[1] (has quite recently had another full assessment published). maybe njalla[2]?
[0] - Iām one of those crotchety fuckers that still has a whole pile of self-hosted things that have been going 15~20y
[1] - seems okay and to have their head on straight. havenāt used myself.
[2] - also havenāt used it myself, comes from some of the folks of the TPB gang
[3] - admittedly I havenāt tried that hard because I donāt need it much, but it is extremely goddamn annoying to debug from clients
I use Posteo for mail and calendar now (theyāre not encrypted between users like Proton but you can just hook it up to any mail client and PGP your shit) .Mail is IMAPS, calendar is CalDAV, contacts are CardDAV, etc. Depending on where you fall on the security-convenience sliding scale, that might be an option. Iāve decided that I care more about portability and standards than super-thick encryption which made me choose them over Tuta, because Tuta offers no way to access the mail over IMAP whatsoever, not even an optional bridge like Proton, and that was a total dealbreaker for me. Posteo also claim theyāre 100% green energy which is a nice bonus.
For drive I use Filen.io now. Theyāre relatively new so I canāt make any assumptions about how long theyāll be around but the price is fair and they offer lifetime payments too. Also their Linux client is pretty solid and doesnāt fucking eat my RAM for breakfast. Theyāre also in the process of adding support for rclone as per a GitHub issue Iām following.
VPN I pretty much donāt use because Iāve never felt I needed it, so no recommendations there from me.
Both of these suggestions are very nice, thank you!
I was in the exact same boat til recently, but switching off of Proton was actually surprisingly easy even though I had it tied into a bunch of accounts and infrastructure. I actually ended up saving a lot of money compared with Proton Unlimited, and itās a relief to not have all my eggs in one basket, especially since stuff like Protonās no logs policy is effectively worthless, and if youāre a whistleblower or similar youāre expected to use a VPN or Tor to access your mail every time to keep from being arrestedā¦ but most likely your VPN (and possibly Tor client) is Proton too if youāre paying for it, with the same worthless no logs policy.
some quick recommendations:
Proton Mail
Proton Calendar
tuta does both of these. their mail is e2e and fine ā itās jankier than proton but also less resource-intensive. itās also the only other choice for now :(
I havenāt used their calendar yet, but from a distance it looks good. I should give it a shot sometime soon.
Proton VPN
this depends on what youāre using your VPN for. actual security? fucked if I know. high bandwidth fuckery? airvpn is pretty good and theyāll let you allocate ports.
Proton Drive
tutaās getting this soon apparently. otherwise, I can second Backblaze being very reasonably priced if you donāt mind having to choose and set up your own e2e software.
Thanks for the suggestions! VPN is mostly to tell my ISP to fuck off. Tuta sounds cool but I am worried about it enshittifying as well. I am relieved to hear that switching from Proton was easy.
Blueskyās approach to using domain names to mean identity is now showing cracks that everyone can see: https://tedium.co/2024/12/17/bluesky-impersonation-risks/
(it was always shaky, but mostly only shown by infosec folks who signed up as amazon s3, etc)
TL;DR: scammer buys .com domain for journalistās name, registers it on bluesky, demands money to hand it over or face reputational damage, uses other fake accounts with plausible names and backgrounds to encourage the mark to pay up. Fun stuff. The best bit is when the sockpuppets got one of the real people they were pretending to be banned from bluesky.
It seems like it is a neat addition to a robust verification system, sadly they picked it as a replacement for a verification system. Ah the libertarian desire to build a thing but not be responsible for it.
this is such a mess, holy shit
and only on .com? I have some very pointed questions about the maturity of the verification program/design
genuine question https://mathstodon.xyz/@sc_griffith/113680680638530340
Consider: making fun of the way they look not only hurts anyone who looks like them through no fault of their own (say, can only afford a shitty cut once every few months) but actively plays into the hands of those shitheels - they made you do a classism.
Letās ridicule them for their garbage beliefs and their shitty actions; they sure deliver plenty of those.
The British elected a guy who wears a mop on his head. You cannot convince me that is his actual hair.
I still have occasional intrusive visions of Johnson busting into an unattended supply closet in the Palace of Westminster to steal a fresh mop head, shouting, āBLOODY LABOUR NICKED ME TOUPĆE!ā
@istewart @techtakes Spoiler: itās not his natural hairstyle, heās been caught deliberately mussing it up before going on camera. Also, his friends and family call him āAlexā (short for Alexander), not āBorisā. Itās all an act.
Heās really sold the bit, hasnāt he?
Yet, if you cover up his hair, he easily scans as āgeneric English aristocrat.ā
And his actual name, Alexander de Pleffel-Johnson, also scans āgeneric English aristocrat.ā
Hence the stage personality.
People probably asked the same question about the first Habsburgs.
mild guess: āgolden boys, with the the 30+ years required to look Politically Evolvedā? and of course the selection factors involved from even just getting to that point and the stylist/image handling that that involves
thereās also an element of the system does as designed, and thereās an element of self-reinforcing delivery/production of these ghouls
Oddly parodied before it happened in the tv series community (the dad of the somewhat racist main character. The dad itself is very racist).
E: I do wonder what Javier Milei looks like if he would dye his hair
somewhat racist
he would continue to look like shit
Briefoverlapping thoughts between parenting and AI nonsense, presented without editing.The second L in LLM remains the inescapable heart of the problem. Even if you accept that the kind of āthinkingā (modeling based on input and prediction of expected next input) that AI does is closely analogous to how people think, anyone who has had a kid should be able to understand the massive volume of information they take in.
Compare the information density of English text with the available data on the world you get from sight, hearing, taste, smell, touch, proprioception, and however many other senses you want to include. Then consider that language is inherently an imperfect tool used to communicate our perceptions of reality, and doesnāt actually include data on reality itself. The human child is getting a fire hose of unfiltered reality, while the in-training LLM is getting a trickle of what the writers and labellers of their training data perceive and write about. But before we get just feeding a live camera and audio feed, haptic sensors, chemical tests, and whatever else into a machine learning model and seeing if it spits out a person, consider how ambiguous and impractical labelling all that data would be. At the very least I imagine the costs of doing so are actually going to work out to be less efficient than raising an actual human being and training them in the desired tasks.
Human children are also not immune to āhallucinationsā in the form of spurious correlations. I would wager every toddler has at least a couple of attempts at cargo cult behavior or inexplicable fears as they try to reason a way to interact with the world based off of very little actual information about it. This feeds into both versions of the above problem, since the difference between reality and lies about reality cannot be meaningfully discerned from text alone and the limited amount of information being processed means any correction is inevitably going to be slower than explaining to a child that finding a āHappy Birthdayā sticker doesnāt immediately make it their (or anyone elseās) birthday.
Human children are able to get human parents to put up with their nonsense ny taking advantage of being unbearably sweet and adorable. Maybe the abundance of horny chatbots and softcore porn generators is a warped fun house mirror version of the same concept. I will allow you to fill in the joke about Silicon Valley libertarians yourself.
IDK. Felt thoughtful, might try to organize it on morewrite later.
Rationalist debatelord org Rootclaim, who in early 2024 lost a $100K bet by failing to defend covid lab leak theory against a random ACX commenter, will now debate millionaire covid vaccine truther Steve Kirsch on whether covid vaccines killed more people than they saved, the loser gives up $1M.
One would assume this to be a slam dunk, but then again one would assume the people who founded an entire organization about establishing ground truths via rationalist debate would actually be good at rationally debating.
Iām guessing these people were JAQing off hard enough that they got kicked out of their local Oxford-style debate org and had to start their own.
Debating post-truth weirdos for large sums of money may seem like a good business idea at first, until you realize how insufferable the debate format is (and how no one normal would judge such a thing).
I mean, the whole point of declaring this era post-truth is that these people have basically opted out of consensus reality.
In other news, Character.AI has ended up in the news again for allowing school shooter chatbots to flourish on its platform.
You want my off-the-cuff take, this is definitely gonna fuck c.aiās image even further, and could potentially leave them wide open to a lawsuit.
On a wider front, this is likely gonna give AI another black eye, and push us one step further to the utter destruction of AI as a concept I predicted a couple months ago.
In the department of not smelling at all like desperation:
On Wednesday, OpenAI launched a 1-800-CHATGPT (1-800-242-8478) telephone number that anyone in the US can call to talk to ChatGPT via voice chat for up to 15 minutes for free.
It had a very focused area of expertise, but for sincerity, you couldnāt beat 1-900-MIX-A-LOT.
FWIW I just got an email from GitHub announcing that Copilot is now free for my account (a very basic one).
Seems like everybody got that email, my account is semi-abandoned and still got it. I love the reek of desperation in the morning
yep I think so too. as I think I posted here a while back:
25068 + Oct 12 GitHub ( 20K) Your free GitHub Copilot access is ending soonand now suddenly itās Launched Again! but with limits. gotta whet those appetites just a bit more! sales will totes follow soon!
mine showed up in spam š¤£
if you think about it the human mind is really just a kind of naturally arising artificial intelligence #Deep
I mean, considering only the relationships between words and symbols in the complete absence of context and real-world referents is a good description of how a certain brand of tech dunce thinks.
The object vs meta level.
