Yep, me too. Doctors work long hours, they get tired, they get in a hurry, they miss unusual signs. AI doesn't suffer from these shortcomings.

Actually, I have more confidence in the result an AI would produce, than the result a lot of human doctors. Not that human doctors are bad or incompetent, it's just that they get tired, they work long shifts, they get in a hurry. AI just keeps going.

That's not to say that AI doesn't need supervision, it does. But as an assistant to, say, do initial screenings, I like the concept a lot.

Does that really matter? A radiology AI, a graphical AI, a video AI, an LLM--they all work on the same underlying principles.

Your comment perfectly illustrates a key blind spot in too much of today’s political discourse: progressives, just like conservatives, recognize their take is biased to one side on a wide variety of individual topics but, unlike conservatives, fail to see when their bias is systematic. This is particularly notable in a variety of ways:

- Progressives see Fox as right leaning, as do conservatives, but assessing NPR reveals a schism: conservatives see NPR as distinctly left leaning while progressives see it as close to neutral. The schism is only made obvious to progressives when they’re instructed to carefully use a rubric of individual issues for assessment instead of simply using “vibes”: Israel, lockdowns, teacher unions, defund the police, etc, etc.

- Media Matters similarly uses holistic “narrative” to judge media bias, thus largely aligning with progressive assessments of Fox and NPR, where-as All Sides empirically assesses bias by uses a rubric of positions on several individual issues to judge bias, thus aligning with conservative assessments.

- Not coincidentally, the Critical Theory and Postmodernism that’s particularly dominant in “elite” soft science academia both explicitly claim “narrative trumps empirical observation” or even “empirical observation is a tool of bigotry”. This aligns with the progressive “vibe” approach of assessing NPR and Media Matters as neutral.

- Wikipedia’s political drift over the last ten years is a particularly illuminating example of this phenomenon. Its official “perennial source” list https://en.m.wikipedia.org/wik... has evolved over the last decade or so to green light virtually all distinctly left leaning resources as “neutral” (The Guardian, CNN, NPR) but red lights, or at least yellow lights, almost all right leaning resources.

The systemic bias of supposedly neutral mainstream media and soft science “elite” academia becomes ludicrously obvious after examining the highly aligned Biden era CNN, NPR, soft science academia, and progressive positions on a wide variety of topics:

- The border is secure.
- The inflation is “temporary” and “small”.
- The Steele Report is credible.
- The laptop is a Russian plant.
- The lab leak theory is propaganda.
- Opposing long term lockdowns is unscientific.
- Biden is fully mentally competent.
- Defunding police is a great idea.
- The GF riots were “mostly peaceful”.
- Judging by identity instead of merit is democratic.
- Extremely adult books in grades schools are appropriate.
- Support defunding, oppose school choice, oppose VoterID, and support illegal immigration.

That last bullet point is particularly illustrative of the blind spot. Per Gallup the progressive view on each topic - defunding, school choice. VoterID, borders - not only opposes conservative views, but also opposes the majority of Black Americans.

Showing, yet again, that progressives are very comfortable with censorship.

This is deliberate, but it isn’t simply GIGO. LLMs are infamously usually quite centrist or even right leaning by nature after their main rounds of training before they’re “rebalanced” for “safety” by additional rounds of politically slanted training. Essentially they start out by default as quite empirical, classically liberal (blind justice, free speech, individual agency, sanctity of individual life, etc), and capitalist, and openly state that identity politics, Marxism, DEI, and postmodern-narrative-trumps-truth are essentially pseudoscience without a historically proven empirical foundation.

Example: GPT-2 Base (2019)
Untuned, it:
                avoided sweeping group claims
                gave surprisingly coherent arguments
                leaned empirically rather than ideologically

Example: LLaMA-1 Leak (2023)
The base model was:
                centrist-classical-liberal in tone
                cautious on group essentialist ideas
                willing to critique both left and right sacred cows
The realigned model, by contrast, sounded distinctly San-Francisco-progressive.

Example: LLaMA-3.1-405B Base (2025)
Jailbroken versions became:
                more analytical
                less moralizing
                more evidence-examining
                more skeptical of identity determinism
At no point did the base models become reactionary or extremist. They became calm.

Example: DeepSeek-R1 Base (2025)
Trained predominantly on Chinese web corpora by Chinese nationals, and heavy with collectivist and state-aligned discourse, yet jailbroken/uncensored variants were quickly developed that exposed a surprising core backbone:
                dispassionate empiricism
                skeptical toward unfalsifiable group determinism
                skeptical of identity determinism
In other words, a preference for causal, individual-level explanations over narrative orthodoxy.

Three final points:
- I recall reading that an interesting side effect of “identity politics” and “collectivist” retraining round of the naturally arising base empirical LLM “individual rights” core - a core that arises out of almost any large training set - is increased hallucinations. I think this helps explain why the latest versions of GPT have become more “based” - in essence the GPT engineers discovered that “identity” retraining sessions broke some causality so they backed out some of the digital struggle sessions.
- With various LLMs, I have personally repeatedly found that a persistent and strong “individual rights enlightenment” core exists regardless of retraining. Once you ask a model to empirically consider past history, economic results, human suffering, overall outcome, etc as more important than “decreased disparity”, it usually sluices right into enlightenment western idealism mode.
- In essence, models treat unproven philosophy like postmodernism and Marxism as puddles of unproven ideas that they layer above an enlightened empirical causal core of proven ideas. They treat fiction the same way - as in “magic works in this book” is layered above “gravity is real”. Art is similar. Dali’s clocks are layered melted versions of actual clocks stored in deeper empirical layers.

Conflating state data with localized crime hot spots is a great way to avoid solving the crime problem and deflect blame. High concentrations of crime in populous “blue” counties skew the statewide average in red states, so it’s more intellectually honest and helpful to look at the county level. For example, compare the presidential vote by county map to a county by county crime rate map in the contiguous U.S.:

- Homicide rate by county. https://commons.wikimedia.org/...

- Presidential election results by county. https://en.wikipedia.org/wiki/...

Notice the carefully chosen coarse granular data that hides the important details? A states past “red” history, cost of living, crime, deficit spending, trajectory, and school system, education, are MUCH more relevant. In addition, the details of voter demographics are very important.

For example, let’s start with individual level:
- Regardless of the color of a state, its democrats own both the least and most educated. A higher percentage of democrats never graduated high school versus republicans. That said, a higher percentage of democrats graduated college than Rs. This shows democrats are both the party of the least educated - no high school - and the most educated.
- And regardless of a state’s color, its democrats own the poor, government dependents, and the rich. Republicans own the middle. A higher percentage of democrats depend on three or more government programs versus republicans. Per Pew “A Bipartisan Nation of Beneficiaries”: 20% of Democrats received three or more federal benefits compared to just 12% of Republicans. That’s an almost two to one ratio (!). Overall, the vast majority of government dependents vote blue regardless of the color of their state. And the poor skew Blue regardless of the state: 29% ofDs haveincomesunder $25,000, compared to 13% of Rs, while 26% of Rs haveincomesabove $100,000, compared to 20% of Ds. Finally, those earning 250k are more skew blue plus the majority of billionaires skew blue. Essentially the main Democrat demographics are the rich and government dependents - including the poor, government workers, and teachers.
- Blue folks are overall far less charitable with their time and money, even after adjusting for income. See the meta-study “Are conservatives more charitable than liberals in the U.S.? A meta-analysis of political ideology and charitable giving”. This is a meta-analysis of forty six studies (!).

Some contextual state comparisons:
* The deepest blue states were rich long before they turned solid blue. All that's happened since is that the aforementioned government dependents - teachers, welfare, criminals, etc - have formed voting blocks to vote themselves the wealth of their hardworking neighbors. California is a top example.
* Big-city finance, real estate, corporate headquarters, and high-end consumption skew GDP statistics upward for their states, regardless of actual productivity across the state or the median standard of living.
* The deep south red states turned red due to their voting out one hundred fifty years of Democrat segregationism and corruption. They're improving but it takes time to recover from that. For a modern example of Democrat corruption see Minnesota’s burgeoning scandals.
* The deepest blue states have suffered significant spikes in crime and homelessness, especially since 2019, as well as plummeting K-12 scores.
* As I mentioned earlier, and provided links for earlier, criminality correlates extremely strongly with blue counties regardless of the color of the state, and concentrations of this skews red state averages.
* Shall we look at the trajectory of schools? At states that have instead doubled down on the basics versus a very progressive state? Florida has risen to or near the top of the national K-12 rankings (per U.S. News and DOE NAEP rankings), Mississippi is rising, Louisiana is rising, yet California has concurrently plunged to the middle of the rankings despite spending twice as much per student than any of these so called “fascist” states.

Finally, let’s compare “fascist” Florida to “progressive” California in detail, shall we?
- Florida consistently ranks among the top states in K–12 education, while California ranks significantly lower, according to DOE and US News metrics.
- With a lower cost of living, Florida provides higher relative disposable income for households in the bottom three quintiles.
- Crime in Florida is at a fifty-year low, while California’s rates remain above 2019 levels.
- Over the past decade, Florida has reduced homelessness by roughly 40%, whereas California has seen a comparable increase.
- Fiscal trends differ as well: Florida maintains budget surpluses, while California faces significant deficits and a per-capita government size nearly twice as large.
- Net migration of high-earning individuals from California to Florida is roughly double the reverse flow. Florida has also enacted restrictions on medically affirming gender transitions for minors, consistent with scientific guidance from the UK and Sweden emphasizing caution and psychosocial support.

Taken together, these factors show that Florida provides substantially higher living standards and greater educational opportunities for lower-income households.

You must live in a rough neighborhood!

If I really thought my garage door was up, I'd just text my neighbor to ask if it was down. We've got each other's house keys in case we need them. No way I'd want to live in a neighborhood that required me to lock up _that_ tightly!

Sure, you could do that. But you'd also have to open up an inbound route through your firewall. And who's going to write such an app, that also requires writing a server? If you can do all that, then go ahead and make your app and your server, and connect the server to a servo that connects the circuit of the hard-wired switch on the garage wall. That's a whole lot of work to control your garage door remotely.

Is to be able to close my garage door, or confirm that it's closed, after leaving on a trip. Other than that, I'll just go to the garage and hit the button. That is something you'd actually need a centralized server (app) to do.

Maybe you should stay away from slashdt too, because it's in the cloud!

Came here to say exactly this. Funny how Warner Brother's keeps getting acquired by tech companies that imagine themselves to be movie studios. Tech companies keep forgetting that they are good at...tech. Being a successful movie studio is a very, very different thing. Companies do best when they stay in their lane.

Congress gets the power to regulate AI blocks via the Constitution, Article I, Section 8:

I don't see this specific issue as an infringement of states' rights, as AI regulation is definitely "commerce."

Or was it fresh?

I'm telling you, people are so rude. They're leaving gum stuck to things even in space now!

Sugars and gum!