I disagree fundamentally. I think any process that uses race as a metric is racist. I don't believe that because a majority of whites was hired in a specific instance that is evidence of racism since, obviously, it could also easily be evidence that in that time, at that place, there were more qualified whites available. That's not especially hard to believe in a population that's already majority-white.

You seem to not know what a straw man is. You need to re-study your logical fallacies

Rochelle Walensky is a real scientist; she is a physician-scientist with a background in infectious diseases and public health, holding an M.D. from Johns Hopkins University and an MPH from Harvard University. In a CBS interview, she stated that "vaccinated people do not carry the virus, don’t get sick," implying no transmission risk from vaccinated individuals. This was based on early data showing low viral loads in breakthrough cases, but the CDC quickly walked it back the next day due to insufficient evidence, acknowledging that vaccinated people could still transmit the virus, though it was "unlikely."

So, stop lying.

So, what? You have reading comprehension problems, genius. I was not implying Myocarditis couldn't be caused by CV19, I was clearly saying folks were denying that young men were at any risk at all from it. That was, in fact, the case. In May 2021 on Twitter, Dr. Peter Hotez referred to myocarditis concerns as "total nonsense" and part of "anti-science aggression," implying reports were fabricated or irrelevant without mentioning infection comparison. He wasn't the only one, either.

“Skeptical" people turn into carriers? Good. I’d rather carry liberty than your strain of obedient, bed-wetting authoritarianism. The real pandemic was millions of cowards catching acute bootlickeritis and demanding everyone else inject the same. Congrats, you just admitted your entire belief system collapses the moment someone says ‘prove it.’ Enjoy your religion, comrade.

That's the problem with lying and politicizing science. Once you lie to folks, even if it's a white lie for a "good" reason, you create doubt that you are ever telling the truth. Lies and exaggerations flew around constantly with a basically worthless/malicious level of political "fact checking" (spin) on anything that emerged. Folks like the CDC or NIH should be above such things, but they played fast-and-loose with the facts and lost a ton of credibility, which probably cost some non-zero number of people their lives. If you want an example that did not come from the media, you can consider how beneficial Fauchi's mask flip-flopping was and justify his initial lying about it.

I know of at least three studies that try to quantify the amount of damage done by constant crawfishing and revisions to "the science" guidance. All of them show major damage done to scientific credibility. This kinda separates the partisans from the truly concerned. If you are a partisan you simply say "Who cares, those damn anti-vaxxers deserved it!" and continue wishing death on your political enemies, just like we saw constantly with pro-government pro-vaccine folks expressing in many surveys that they found the idea of putting people in concentration camps or taking anti-vaxxer children to be a-ok great ideas. Someone who is really concerned about saving lives would say "Wait, we need to preserve the credibility of scientific guidance!" I understand that some of the science was dynamic and emerged in real-time, but other things like mask-guidance, should have been a no-brainer.

Sorry, I don't speak moron.

Your example plays exactly into the scenario I mention. We can see the core motivations of these schemes and dubious "concerns" when they get prescriptive. Somehow the "fix" is always "hire fewer whites" and they do the same hand waving "but but but, don't worry, it's not a zero sum game, we are just giving that job to a minority, not because we are anti-white or anti-male though, never that. If we were that'd be justified, though." That's an extremely suss position and that's exactly the position Google was taking.

Requoted against the censor trolls with mod points. I should ditto several following related comments, but Slashdot isn't worth that much effort these years.

My top candidates just now:

1. It's just a joke.

2. I'm just asking questions. (Most relevant to this story.)

3. AI is good.

So what's your favorite?

In my typically verbose way, I feel like a few words of clarification are called for. Also another attempted joke or two?

The first one is mostly frequently abused as an excuse for bad behavior, including speech behaviors. In particular, there are many lies that used to be taken as proof of character flaws, but now they are just spun away. In orange particular, "The president was only joking" is no excuse for a job that ain't supposed to be so funny it makes you sick. (Which actually comes back to the theme of the Slashdot story at hand.)

The second one is most damaging as an epistemological attack on the nature of truth itself. It's actually a good thing that science does ask questions, but the goal of scientific questions is to learn more, not to destroy the idea that we know anything at all. Perfect knowledge should not be the ultimate enemy of trying to learn anything at all on the excuse that our knowledge ain't perfect. As if there were any perfect scientists (or politicians), now or ever.

Now about my newish third candidate, the problem is with "good". Options that are closer to the truth might be "AI is a tool too easily used as a weapon" or even "AI is nothing" because it's the human beings who use things, even including AI things.

Just had another encounter with an AI entrepreneur yesterday. Language-related application should have caught my interest, but his money-centric attitude lost it. My bad. What else should I have expected at a VC gathering? The main reason he was there was in hopes of getting some of that sweet, sweet cash and I should congratulate him on his tight focus. (A-hole joke time?)

Back to the AI threat. I suppose the main angle for this story should be examples of AI slop attacking vaccines in particular and the CDC in general. Too depressing to websearch for some examples, and you can get AI help if you want some. I'm more focused on the GAIvatar threat. I considered "GAIvatar are harmless" as my third candidate, but the portmanteau is not frequent and I've been unable to find any standard usage describing generative AI used to imitate specific people. Rarely they may offer a few bits about chatting with a fake Einstein or an AI ghost of a grandparent. Recently read an interesting SF story about solving a major math conjecture with the aid of an AI postmortem copy of a deceased father...

So I used to focus on the use of individual GAIvatars to predict and control individuals (though carefully crafted and targeted prompts). But now I'm wondering about creating a group GAIvator to predict and control the behaviors of an entire class of people. It could even become a kind of circular definition, where group membership is defined on a sliding scale based on how closely a particular candidate member conforms to the GAIvator's predictions and prompts.

So have a nice Friday?

Me? I'll take my chances with the vaccines. Much better odds than they'll give me in Vegas or the stock market.

Heard that a lot along with "the vaccine will stop transmission" and "you're imagining young men getting myocarditis" and "Children are at severe risk for COVID complications. Vaccinate them all like Pokemon." The trouble with your "science" is that, to you, it's some kind of left-wing religious term. To me, it means looking at the actual data and staying skeptical, which clearly .... you didn't. Your version of "science" had idiots out dumping sand in outdoor skate parks so the kids would crowd together to play video games indoors.

*YAWN* I love the smell of jealousy in the morning.

Well, suicide rates are strongly linked to many factors such as geography and weather. Fertility tends to be the highest in the biggest poor shitholes, regardless of their government. Let's examine something voluntary like where people actually want to live. Do we see that people migrate toCommunist countries or away from them? The evidence there is clear: given a choice, people want freedom and small government. Nobody is fighting to get into Cuba or North Korea despite being Communist paradises.

Yeah, a $773,000 deal soured into breach-of-contract claims, exposing the folly of over-engineered e-gadgets and proprietary DRM that locks out repairs and dooms machines to early graves, exactly what most farmers HATE.

Enter the unsung heroes: venerable tractor makers like sometimes Ford and International Harvester, whose mechanical marvels shun "stupid electronic shit." These workhorses, think the pre-DRM 1960s John Deere 4020 or the indomitable IH 1066, don't have DRM shackles, letting farmers wield wrenches freely. No subscription-locked software, no fleet managers spying via GPS; just pure, repairable engineering built to outlast owners. That's why you see these tractors being passed down in families

And farmers are voting with their wallets. The used tractor market, projected to swell from $59 billion in 2025 to nearly $75 billion by 2035 at a 2% CAGR, is a gold rush for these old tractors. Sales of pre-owned ag gear are eyeing 18% growth this year, with 60% of operators eyeing upgrades to proven vintage iron over shiny new failures. Yeah, that's right: Longevity, baby. A well-maintained traditional tractor clocks 4,000–15,000 hours, or 15–30 years of sweat equity. Some Midwest beasts from the 1920s still plow fields today, century marks in sight. Many many Ford 8N's are still hauling hay at 80+. John Deere's 4440 has many of 20,000-hour runs.

In an era of planned obsolescence, these dinosaurs endure because they're farmer-proof: simple, souped-up with aftermarket parts, and free from corporate kill-switches. As used auctions hum and old-timers rumble on, the outcome is clear: in the dirt, dumb and durable beats smart and fragile every time.

The battle cry of just about every Communist I've ever talked to. They don't talk for long, though. They usually have a hunted look in their eyes, worried you'll actually know the history of Marxism and Communism and will bring up it's many mass-homicidal failures. Then it instantly degenerates into "Well.. Hmph. Pfft. Grrr. That wasn't REEEEEAL communism."

I've always wondered what an OS would look like based only on power-users' feedback. Like the number of licks to get to the center of a Tootsie-Pop: the world may never know. That's why I feel "The Unix Way" (the 9 principles laid out best by Mike Gancarz) was so awesome. It tried to take the best ways of doing things and merge them into a coherent operating philosophy. I still haven't seen a better paradigm.

A key limitation I’ve observed isn’t only the scope or quality of training data, but the model’s fixed context window (i.e., maximum token limit). Once a codebase exceeds roughly 300–400 lines (depending on the model and tokenizer), earlier portions of the code, prior instructions, architectural constraints, or critical logic fall outside the current context. As a result, the model effectively “forgets” them and can no longer reason coherently about the program as a whole.

This makes LLMs reasonably effective for short, self-contained scripts, where they can occasionally produce elegant or creative solutions, but fundamentally unreliable for large, complex codebases. The model simply cannot maintain global consistency or juggle all relevant constraints simultaneously because significant parts of the program are no longer in its active context. In essence, it runs out of working memory, not unlike a programmer trying to hold an entire 10,000-line system in their head at once. Programmers still have compressed summary in their heads of the overall goals & previous code. LLMs just blithely spout code without that benefit and it shows.