Stubsack: weekly thread for sneers not worth an entire post, week ending 19th January 2025 - awful.systems
Stubsack: weekly thread for sneers not worth an entire post, week ending 19th January 2025 - awful.systems
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 you’ll 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 cut’n’paste it into its own post — there’s no quota for posting and the bar really isn’t that high.
The post Xitter web has spawned soo many “esoteric” right wing freaks, but there’s no appropriate sneer-space for them. I’m talking redscare-ish, reality challenged “culture critics” who write about everything but understand nothing. I’m talking about reply-guys who make the same 6 tweets about the same 3 subjects. They’re inescapable at this point, yet I don’t see them mocked (as much as they should be)
Like, there was one dude a while back who insisted that women couldn’t be surgeons because they didn’t believe in the moon or in stars? I think each and every one of these guys is uniquely fucked up and if I can’t escape them, I would love to sneer at them.
(Semi-obligatory thanks to @dgerard for starting this.)
Remember how OAI claimed that O3 had displayed superhuman levels on the mega hard Frontier Math exam written by Fields Medalist? Funny/totally not fishy story haha. Turns out OAI had exclusive access to that test for months and funded its creation and refused to let the creators of test publicly acknowledge this until after OAI did their big stupid magic trick.
From Subbarao Kambhampati via linkedIn:
"𝐎𝐧 𝐭𝐡𝐞 𝐬𝐞𝐞𝐝𝐲 𝐨𝐩𝐭𝐢𝐜𝐬 𝐨𝐟 "𝑩𝒖𝒊𝒍𝒅𝒊𝒏𝒈 𝒂𝒏 𝑨𝑮𝑰 𝑴𝒐𝒂𝒕 𝒃𝒚 𝑪𝒐𝒓𝒓𝒂𝒍𝒍𝒊𝒏𝒈 𝑩𝒆𝒏𝒄𝒉𝒎𝒂𝒓𝒌 𝑪𝒓𝒆𝒂𝒕𝒐𝒓𝒔" hashtag#SundayHarangue. One of the big reasons for the increased volume of "𝐀𝐆𝐈 𝐓𝐨𝐦𝐨𝐫𝐫𝐨𝐰" hype has been o3's performance on the "frontier math" benchmark--something that other models basically had no handle on.
We are now being told (https://lnkd.in/gUaGKuAE) that this benchmark data may have been exclusively available (https://lnkd.in/g5E3tcse) to OpenAI since before o1--and that the benchmark creators were not allowed to disclose this *until after o3 *.
That o3 does well on frontier math held-out set is impressive, no doubt, but the mental picture of "𝒐1/𝒐3 𝒘𝒆𝒓𝒆 𝒋𝒖𝒔𝒕 𝒃𝒆𝒊𝒏𝒈 𝒕𝒓𝒂𝒊𝒏𝒆𝒅 𝒐𝒏 𝒔𝒊𝒎𝒑𝒍𝒆 𝒎𝒂𝒕𝒉, 𝒂𝒏𝒅 𝒕𝒉𝒆𝒚 𝒃𝒐𝒐𝒕𝒔𝒕𝒓𝒂𝒑𝒑𝒆𝒅 𝒕𝒉𝒆𝒎𝒔𝒆𝒍𝒗𝒆𝒔 𝒕𝒐 𝒇𝒓𝒐𝒏𝒕𝒊𝒆𝒓 𝒎𝒂𝒕𝒉"--that the AGI tomorrow crowd seem to have--that 𝘖𝘱𝘦𝘯𝘈𝘐 𝘸𝘩𝘪𝘭𝘦 𝘯𝘰𝘵 𝘦𝘹𝘱𝘭𝘪𝘤𝘪𝘵𝘭𝘺 𝘤𝘭𝘢𝘪𝘮𝘪𝘯𝘨, 𝘤𝘦𝘳𝘵𝘢𝘪𝘯𝘭𝘺 𝘥𝘪𝘥𝘯'𝘵 𝘥𝘪𝘳𝘦𝘤𝘵𝘭𝘺 𝘤𝘰𝘯𝘵𝘳𝘢𝘥𝘪𝘤𝘵--is shattered by this. (I have, in fact, been grumbling to my students since o3 announcement that I don't completely believe that OpenAI didn't have access to the Olympiad/Frontier Math data before hand.. )
I do think o1/o3 are impressive technical achievements (see https://lnkd.in/gvVqmTG9 )
𝑫𝒐𝒊𝒏𝒈 𝒘𝒆𝒍𝒍 𝒐𝒏 𝒉𝒂𝒓𝒅 𝒃𝒆𝒏𝒄𝒉𝒎𝒂𝒓𝒌𝒔 𝒕𝒉𝒂𝒕 𝒚𝒐𝒖 𝒉𝒂𝒅 𝒑𝒓𝒊𝒐𝒓 𝒂𝒄𝒄𝒆𝒔𝒔 𝒕𝒐 𝒊𝒔 𝒔𝒕𝒊𝒍𝒍 𝒊𝒎𝒑𝒓𝒆𝒔𝒔𝒊𝒗𝒆--𝒃𝒖𝒕 𝒅𝒐𝒆𝒔𝒏'𝒕 𝒒𝒖𝒊𝒕𝒆 𝒔𝒄𝒓𝒆𝒂𝒎 "𝑨𝑮𝑰 𝑻𝒐𝒎𝒐𝒓𝒓𝒐𝒘."
We all know that data contamination is an issue with LLMs and LRMs. We also know that reasoning claims need more careful vetting than "𝘸𝘦 𝘥𝘪𝘥𝘯'𝘵 𝘴𝘦𝘦 𝘵𝘩𝘢𝘵 𝘴𝘱𝘦𝘤𝘪𝘧𝘪𝘤 𝘱𝘳𝘰𝘣𝘭𝘦𝘮 𝘪𝘯𝘴𝘵𝘢𝘯𝘤𝘦 𝘥𝘶𝘳𝘪𝘯𝘨 𝘵𝘳𝘢𝘪𝘯𝘪𝘯𝘨" (see "In vs. Out of Distribution analyses are not that useful for understanding LLM reasoning capabilities" https://lnkd.in/gZ2wBM_F ).
At the very least, this episode further argues for increased vigilance/skepticism on the part of AI research community in how they parse the benchmark claims put out commercial entities."
Big stupid snake oil strikes again.
Starting to think we're about at the point where you could make the best search engine on the market in these three easy steps:
- Search Wikipedia for whatever the user typed and show the top result first.
- Check if <search keyword> dot com, org, and net exist and show them in the order of popularity.
- End of page.
(oh no it's politics)
Trump's new cryptocurrency scheme is surprisingly forthright about being a pump & dump:
CIC Digital LLC, an affiliate of The Trump Organization, and Fight Fight Fight LLC collectively own 80% of the Trump Cards, subject to a 3-year unlocking schedule. CIC Digital LLC and Celebration Cards LLC, the owners of Fight Fight Fight LLC, will receive trading revenue derived from trading activities of Trump Meme Cards.
Essentially according to their own website, they started by selling 20%* of the tokens to the public, and over the next few years will... sell another 80% of the tokens to the public. To the moon!
* half of that they describe as "liquidity" instead of public distribution -- whatever that means.
I read about this gross Robo Anne Frank LLM by a company called "School AI": Bluesky post (looks like via an activitypub bridge, but I can't be bothered to find the canonical link), News Article, School AI's website.
Gee it sure is weird how all these digital clones the AI companies keep coming up with all have the exact same (lack of a) personality.
So, the Wikipedia article about "prompt engineering" is pretty terrible. First source: OpenAI. Second: a blog. Third: OpenAI. Fourth: OpenAI's blog. ArXiv, arXiv, arXiv... 43 times. Hop on over to the Talk page, and we find this gem:
It is sometimes necessary to make assumptions to write an article (see WP:MNA).
Spoiler alert: that link doesn't justify anything. It basically advises against going off on tangents: There's no need to rehash the fact that evolution is a fact on every damn biology page. It does not say that Wikipedia should have an article on some creationist fantasy, like baraminology or flood geology, based entirely on creationist screeds that all cite each other.
hells yeah it's time for some action - Drew DeVault is organizing a sit-in protest of Jack Dorsey's keynote at FOSDEM 2025.
okay so I’ve just found a new shirt I want
there it is, sammy has gone and said people are just prompting the model wrong (I recall we’ve had that bit said here earlier)
but in true sammy grift: you just need to be asking the right questions to trump intelligence. “why do you want to suck, as a human?” sammy asks, not understanding a moment of humanity
This is extremely tangential to the areas of sneer interest, but seeing as this is the only technology related community I am in, I’m putting it here.
This song has been making the rounds on the charts/social media and I refuse to believe that it isn’t about the package management tool apt
some of the first research science on promptfondlers and model-affine dipshits is starting to see the light of day and, in what will surprise probably 0% of our regulars, it confirms some things
(I have grumped about their desire for outsourced thinking in the past myself)
a couple weeks back, I was (bc reasons) looking around to see how to turn off goog's annoying gemini bullshit in an account, and you can!
except then even after doing that, accounts in that org still got prompts (in the form of in-app banners, and sparklebuttons in shit like gmail) to Try The Model
it looks like people aren't biting enough, because now you get it whether you like it or not, for the low low price of pushing up your base account fee! and I checked in one org - "Gemini App" is disabled org-wide, but the fucking prompt is immediately in the UI (and you get a modal popover opening gmail)
fuck these people so much
Some news on the people who own/run protonmail. They are pro Trump
Possibly I’m the last to hear about this one, but seeing as proton mail has come up here a few times before: the founder and ceo Andy Yen is apparently a Trump fan.
Great pick by @realDonaldTrump. 10 years ago, Republicans were the party of big business and Dems stood for the little guys, but today the tables have completely turned. People forget that the current antitrust actions against Big Tech were started under the first Trump admin.
(from the beginning of december, on the nomination of trump staffer Gail Slater to antitrust post at the doj)
Scooter pulls out Richard Lynn’s national IQ figures. https://archive.is/BJ6bd
new developments in things understanding by promptfondlers https://feddit.uk/post/22751391
Free SFnal short story idea, came to me literally in a dream:
Dude is living his best life, beatiful house, beatiful wife, gets a job doing computer stuff "improving the world". But his big fancy work computer is wasting a lot of space so he reformats it/installs Nix, and suddenly everythings gone, all grey wireframe, no way out. Turns out he was given root to his own simulation and there's no backup.
Feels I should have read this somehwere but haven't read short SF in ages so...
With risk of falling into the 'classify people into two binary groups' thing which I have often criticized the Rationalist for. Move over jock vs nerd. There is Jock vs Creep.
Computer understanders, click this link for some psychic damage! :D (aside: I tried to upload this as an image but it didn't work)
nasb, it warms my cold diamond heart to continue seeing instances of people going "oh, yeah, uh.... actually, no, fuck that shit. we'll do it ourselves" in response to broadcom's attempt to Raise Revenue Through Product Capture
