Big brain tech dude got yet another clueless take over at HackerNews etc? Here's the place to vent. Orange site, VC foolishness, all welcome.
This is not debate club. Unless it’s amusing debate.
For actually-good tech, you want our NotAwfulTech community
Hi, I'm new here. I mean, I've been reading but I haven't commented before.
I'm sure you all know about how cheap labour is used for labelling data for training "AI" systems, but I just came across this video and wanted to share. Apologies if it has already been posted: Training AI takes heavy toll on Kenyans working for $2 an hour.
Welcome! The situation these people find themselves in is dire, since they're both literally and emotionally as far as possible from the people making decisions about their labor. The modern economy doesn't function without exploitation, and generative AI is the latest innovation in expanding that exploitation and pushing it farther away from the people who benefit and who make the decisions that require it. It does to modern knowledge workers what automation and outsourcing did to manufacturing, and the distance is sufficient that I don't expect to see even the kind of lukewarm pushback that sweatshops got in the 90s actually manifest for them.
In the aftermath of an LGBT hate incident, the then-CEO of cloud computing giant Digital Ocean told upset staff his mentor was a member of the KKK as an attempt to explain why they must bend their values because "we love the company"
presented without comment
It somehow gets worse. These people think values are interchangeable with value.
Asking employees to "bend" their perfectly sensible values like "I don't like homophobes" or "members of the KKK suck" is insane to me, but exactly the sort of thing a tech CEO would think would resonate with his workers.
I stay at my job not because I have molded my soul into a perfect vessel for my companies values (which, TBH, kind of suck), but because I have a mortgage payment.
(Also as the header graphic points out, "love is at our core" and "inclusive environment" are apparently some of their values so maybe it's Digital Ocean which needs to bend to Digital Ocean's values).
At least there's a happy ending:
A month after the all-hands meeting, in August 2023, DigitalOcean announced that it was conducting a search for a new CEO, but did not say why.
edit: context https://www.independent.co.uk/tech/chatgpt-david-mayer-name-glitch-ai-b2657197.html
Time for another round of Rothschild nutso's to come around now that ChatGPT can't say one of their names.
At first I was thinking, you know, if this was because of the GDPR's right to be forgotten laws or something that might be a nice precedent. I would love to see a bunch of people hit AI companies with GDPR complaints and have them actually do something instead of denying their consent-violator-at-scale machine has any PII in it.
But honestly it's probably just because he has money
I think Sam Altman's sister accused him of doing this to her name awhile ago too (semi-recent example). I don't think she was on a "don't generate these words ever" blacklist, but it seemed like she was erased from the training data and would only come up after a web search.
Being erased from the training data is frankly even more galling than the kind of brute force GDPR compliance they seem to have been using. It puts the lie to any claim that they're just "moving fast and breaking things" without mind to the consequences, because clearly there was some reason to prune the training data and make sure that the model didn't have certain information when it was to the company's (or the founder's) liking.
My hands are requesting the CEO's home address (in ~~Minecraft~~ real life):
Quick bonus I found in the replies:
And a quick sidenote from me:
This is sorta repeating a previous prediction of mine, but I expect this AI bubble's gonna utterly tank the public image of tech as a whole. When you develop a tech whose primary use case boils down to "make the world worse so the line can go up", its gonna be virtually impossible for the public to forgive you.
Being more specific, I expect artists/musicians/creatives in general to be utterly hostile to AI, if not tech as a whole - AI has made their lives significantly harder in a variety of ways, and all signs pointing to the tech industry having done so willingly.
really into this whole far right coup and far right right hegemony thing a lot of countries are trying out rn. really cool
As resources get scarcer the lurch back to fascism is going to get worse. People always turn to the simple (but categorically wrong) answers when times get hard, and blaming people who come from other places has historically been a winner for simple and wrong answers. Refugees are a symptom but they'll be targeted as the cause because their numbers will swell with increasing "natural" disaster frequency and severity.
And I have no idea what I can do about it that wouldn't amount to pissing in the wind.
Sometimes I wonder what future generations will think looking back on the content of today's internet and humanity's total inability to demonstrate the humane.
Looks like SK gave the bastards a hard nope, at least.
this was the most half-assed coup in recent memory, hands down
hey army do a coup for me
parliament said no
ok
no fanatical loyalists in army, no militia or PMC, zero planning at all, even Prigozhin got that sorted out. very similar thing happened just two years ago in Peru, and this year in Bolivia. and same thing will happen this time, impeachment is already written as we speak. army also exactly knew what was going on and didn't participate for a reason
by the way, south korean presidents don't have a habit of leaving office peacefully, so it all tracks. it's either impeachment or coup or long prison term just after leaving office
An update, a user stated, "with all love and respect": "people will not starting hating Brave but they will start to hate the mod who blocked Brave".
2 hours later they'd removed their account.
Older article, but relevant in light of current events:
UnitedHealth uses faulty AI to deny elderly patients medically necessary coverage, lawsuit claims
I wonder if we'll see a new edition of the Shinzo Abe Effect.
also, tech billionaires terrified of a peasant uprising can turn their homes into fucking military bases with a home security system named (of course) Sauron
LOTR fans stay losing
Fuckers like Thiel make me want to start stuffing guys who take too much interest in speculative fiction into lockers again, just to be safe. And I was that kid, right down to the STEM focus and disrespect for the liberal arts.
It's never wrong to gatekeep fascists.
Sorry about more insta content. Saw this today and it tickled me:
Left: pinker
Right: Yeonmi Park, who is known for defecting from north korea and speaking about it. That should be enough if she just told the truth, but for one reason or another (grifting dosh) she has started just making shit up, which I recommend you research for yourself if you are interested.
Anyway the above picture probably doesn’t come as a surprise when you consider her current tack: saying that “woke” and “cancel culture” are the first steps in a country becoming like North Korea. She’s been on JRE (where she memeably claimed that there is only one train in NK that often requires people to push it). She has a book with her anti-woke thesis with opening shit-volley by old mate JBP.
Man that page has got some really wild sentences.
Park's claims that human corpses commonly float down North Korean rivers were criticized by Swiss businessman Felix Abt, who lived and worked in the DPRK for seven years.
Now I don't know if there's any validity to her claim, but love the expert testimony from Swiss entrepreneur who's been allowed to tour Pyongyang for a bit.
It's frustrating because there is plenty of real evidence that North Korea is in a bad way. That satellite image is stark and the volume of reports claiming food shortages and lack of basic services are hard to dispute, as is the sheer obsessive control exercised by the government to avoid the depictions of it. You don't need to make up stories about using human corpses to fertilize your fields or whatever.
Ah, yes. In case you were wondering about said Swiss businessman and whether or not his actions in the DPRK were morally sound, once again, Wikipedia has something to say.
Dont think that anybody involved in this is going to be a good person. Of all the people mentioned, Pinker is prob the least worse here. And that is saying much.
Somehow not the most horrific thing Swiss did just by trading (that would be war profiteering during Iran-Iraq war, when they sold chemical weapons precursors to one side and gas masks to other)
it's almost as if international trade of weapons creates perverse incentives
LOL @ North Korea being "woke". If they are, why are they assisting Russia, the Right's Great White Hope, in a war against a country that these people also denigrate as "woke"? But hey, never let facts or logic get in the way of a good grift.
I mean, with how young she is there's absolutely no credibility she can offer on how a country becomes like the "D""P""R"K. But I can see the superficial similarities between the right-wing griftoverse's depiction of cancel culture and the kind of cultural authority that gets exercised. That Bogeyman is why they use that fanciful description, after all. I can see how easily she could get radicalized from there, not that it justified any of the lies she's believed and passed on or invented outright.
And right now South Korea's president has declared martial law and banned all political activity, I guess she'll be just fine with that.
Martial law and free speech restrictions are only bad if the bad guys do it, also we get to decide who the bad guys are
That's what she said? That's literally so much funnier than any of the memes
i read the title and was like damn we're dunking on game engines now?
for a post from an AI company this is surprisingly honest about the general helplessness in defeating so called hallucinations
https://www.kapa.ai/blog/ai-hallucination
probably a good sign for even boosters to be talking this way
The first half was OK, but then they cited this paper.
LLMs encode much more information about truthfulness than previously recognized. We first discover that the truthfulness information is concentrated in specific tokens, and leveraging this property significantly enhances error detection performance. Yet, we show that such error detectors fail to generalize across datasets, implying that—contrary to prior claims—truthfulness encoding is not universal but rather multifaceted.
I haven't read the paper, and probably won't, but what the shit is this?
gave it a quick skim. I lack any relevant background. the bit they push most seems to be that you can improve the performance of error detection tools by determining the most important tokens in an answer and running your tools on the tokens near those. this seems to be in contrast to absurdly naive approaches like averaging the tokens (???) or just looking at the last token of the response (????).
what are the most important tokens? they're the ones that change the factuality of the answer if you change them. how do you determine that? you don't, lmao. you just ask an LLM what the most important words are
what are the error detection tools? you will never guess
It turns out if you can just make the machine know what the truth is and say that you don't get hallucinations. Unfortunately the truth isn't emergent from pure language models and expressing Truth through language alone has been something challenging the human race since Krog try to teach Torg how make stick but pointy.
a good follow on insta if you use it
That guy must be a communist. Cause that is a big red flag.
