Terr_

So, uh, there's something very off here, and I just can't parse WTF the author is trying to claim or convey. Is there missing-context, like it being a follow-up to a prior post? Was there some formatting that broke down so that it's no longer clear what's a quote or from whom? Perhaps it is just some stream-of-consciousness that needed to cook a more before public consumption.

My most-charitable interpretation:

1. There is some kind of vendor or employee tasked with transcription, and implied to be paid through public tax-money.

2. They cited or billed 4h for a speech, but delivered shit with errors in it in just 5 minutes, presumably using an LLM.

3. Then the author became angry over it but didn't quite tell the story.

P.S.: I want to make it clear that this isn't just about dealing with other humans. The dread I contemplate should affect even introverted shut-ins.

I want to reduce accidental complexity, discover and understand real problems, solve them well, and leave the world somehow better for it. Not just some infinitely process ping-pong of a hallucinated digital dollhouse fifteen levels removed from objective reality.

I don't think anyone's talking about towns where electricity is absent or expensive.

Instead it's locations that begin with cheap electricity (attracting the datacenter) but the residents fear a significant increase in electricity costs with no local economic benefits to offset it.

I have a premonition of future programming despair, and I wonder if it'll be like what Ian McKellen felt: A veteran of Shakespearean plays, doing all those [correction, not LoTR] The Hobbit green-screen shots interacting with no real person...

__________

McKellen said: "I felt pretty miserable … and thought perhaps, has the time come for me to stop acting altogether if I can't cope with these difficulties?" [...]

"It was so distressing and off-putting and difficult that I thought 'I don't want to make this film if this is what I'm going to have to do'," McKellen added. "It's not what I do for a living. I act with other people, I don't act on my own."

-- https://www.theguardian.com/film/2013/nov/20/the-hobbit-gand...

I am assuming that is the point you are trying to make.

Not exactly. My point is that LLMs dramatically increase the danger and risk of certain types of manipulation, fraud, and brainwashing. Those things will become much easier to do cheaply and secretly and at a large scale.

The Truman Show offers a useful way to convey those kinds of risks, but the details of the product being sold doesn't really matter, we just need to know that the company is willing to do those kinds of things.

no such land formation

I'm sure the script-writers knew this. I believe they chose it because they wanted to make a satire (mocking joke) of real-world coffee-products being sold at a certain time, but did not want to include details that that some real-company (or country) could get too angry about.

I want to raise the possibility that you (pixl97, simoncion) actually hold many of the same opinions but are clashing because of a different definitions the LLM / bad-thingy scope.

* Narrowly - The core algorithm that extends documents cannot be made safe, because it's a stochastic machine with no data/instruction separation possible. Unanticipated input can evoke arbitrary output.

* Broadly - The overall offering (centered on the document-extender algorithm) could be made safe by limiting its over-ambitious scope, treating the document-extender output as malicious-by-default, and sharply limiting what that output can drive or influence. Of course, that would exclude the berjillion-dollar stock valuation replace-all-humans stuff.

LLMs are impossible to make safe in the same sense that humans cannot be made safe.

I am very conflicted by this sentence, the two halves being:

1. Yes, the futility of making LLM's "safe" in that rigorous way is insurmountable, barring a major algorithm rewrite, and nobody really knows what that could be yet. Anyone who says it's easy is glossing over details--or selling something.

2. No, the failure modes of LLMs are substantially different than humans. If someone thinks they're similar, then they will fail at estimating and containing the risks. Now, perhaps if the comparison was to a brain-damaged human hopped up on psychedelic mind-altering drugs...

Note that I'm distinguishing here between the LLM itself--the hyper-mad-libs story generator--versus regular programs around it.

I think that's an equivocation, which blends two extremely different kinds of "dangerous", ex:

1. "Our new car has soo much raw power and incredible armor on it, be glad we're the ones building or else bad guys would use a fleet of them to take over the world! How will you stay safe without being in one yourself? Invest today or be left behind!"

2. "So, uh, nobody can consistently steer our car properly, it keeps veering sideways sometimes, especially at high speeds, and people are finding sneaky ways of tricking it into slamming into barriers and turning pedestrians into pink fog..."

What's left is the code alone.

The necessary work is more than just the deliverable. Even a simple patch implies an indefinite number of other approaches not-taken, and sometimes not-taken for good reasons.

Sometimes we explicitly document those, sometimes we trust that a human expert decided against them. Those other aspects of work still exist, even if they are getting skipped/mismanaged with LLMs.

It’s the same with writing: AI writing is coherent on the surface, but is impossible to edit because it’s built on no real insight.

LLMs lie in the unhappy-medium between an abstract machine we can reason-about versus a person we can instinctively model and simulate.

Instead it's a complicated machine that evades both reasoning and intuition.

At the risk of stating the obvious: We've ample evidence for this insidious effect with ~25 years of Google changes, as clear ads become hidden ads, separation becomes deliberate clutter, and actionable results get degraded into interaction-bait.

Not the source I wanted to recall/find, but a timeline with screenshots:

https://www.howtogeek.com/how-google-search-results-have-cha...

I'm quite conscious of being repetitive here, but The Truman Show is relevant again:

Why don't you let me fix you some of this new Mococoa drink? All-natural cocoa beans from the upper slopes of Mount Nicaragua. No artificial sweeteners. I've tasted other cocoas. This is the best.

Lost Coast [...] This is also how your pupils work, of course.

As a purely temporal change it was neat, but I recall being distinctly annoyed with Lost Coast over how that effect made things worse (less realistic) for scene elements not under the crosshair / gun / avatar's-nose. We naturally expect contrast (and focus, see DoF rant) to improve when we direct our auto-adjusting eyeballs towards the area. In other words, realism of moving between light-environments came at the expense of realism in glancing around the room.

The straightforward fix would be gaze-tracking hardware... but now that I think about it (and with 20 years of tech improvements) I can imagine a much more mad-science approach: Track a simulated visual memory for the avatar!

Imagine a game that tracked recent virtual saccades to build up a cache of data like gaussian splats. So if that player has already seen the black notebook on the black desk in the dark room, it shows up more easily even if they walk in and out of the bright outdoors.

They are being sarcastic.

I think you're conflating two separate issues:

1. The accuracy of the label.

2. The likelihood the label will cause problematic misunderstandings.

When my rice-cooker logic is advertised as "AI", that's a stretch, sure... But it's extremely unlikely to cause an investment bubble seeking the Rice Cooker Economic Singularity, incur protests from the Rice Cooker Emancipation League, or lead to weird folks in their basement seeking divine wisdom from its vaporous whispers.

Good luck timing it. "The market can remain irrational longer than you can remain solvent."

There isn't always a mechanism that will translate being-right about some fundamental fact into making-money from it.

P.S. It isn't trivially "solved" by someone setting up a gambling scheme either. Many people who are right will still lose because they didn't also guess when other people would start to agree with the conclusion. Plus there's the time-value of money for things which take longer to play out.

An AI bailout would be a poison pill that would electorally doom whichever party was behind it for decades.

Two concerns:

1. In a duopoly it won't doom either one party, not if the bailout-seekers involved are able to bribe them both equally.

2. A politician in not-so-bad party could let the other party doom itself in a long-term diffuse sense... Or they could accept some generous short-term very concentrated-on-me donations. The interest of the group is not always the interest of the members.

The US desperately needs voting-reform, to reduce the spoiler-effect and winner-take-all mechanics.

ACLU credibility lost with such a sensationalist headline.

HN Comment credibility lost with such a sensationalist one-liner?

I see your claim, but I don't see how you got there. TFA has enough examples that I don't think you can just casually pooh-pooh everything away without an argument.

Seems only fair

"You're trying to kidnap what I've rightfully stolen!" -- Vizzini

Right, too many games pitch the "realism" of "really like a Hollywood movie", reproducing effects which are really telltale flaws in cameras, microphones, or the theater-experience.

This is in contrast to the realism of "if you were there in person". If you step outside and look towards the sun and see lens-flares, consult an eye-doctor. If you see polyhedral lens-flares or bokeh... consult a roboticist.

I'd like to highlight that even if one added an URL-exists step [0], that doesn't do a dang thing for result-set problems of:

1. False-negatives, where relevant posts that do exist are not being shown (imagined or otherwise) to the user.

2. Posts which exist but don't fit the words the chaos-parrot uses to describe them.

3. "Relevance" being determined by unpredictable factors that aren't stable, predictable, or desirable.

In other words, it's just more whack-a-mole lipstick-on-a-pig third-animal-idiom-here.

[0] A bad idea on its own, since it creates a security vulnerability for data-exfiltration or indirect malicious attacks.

I see this one under "new" (2 min ago) but no other submissions or comments in your profile. (?)

Condensed rant: Humans are blind to the most common gasses on our atmosphere (N2, O2) and sometimes have difficulty perceiving its most common liquid (H2O). This is sensible, since otherwise we'd be helplessly wandering through "visible" fog to fall off cliffs or be eaten by giant carnivorous bats. Our "colors" are mostly a quirk of where we're from.

Now suppose some aliens arrived at our planet and said: "Wow, your planet is the most-boring we've ever seen, it's all just one N2-colored ball." We'd be a little offended! Clearly, the dumb aliens are looking at it wrong!

Flip that in reverse, and we have an artistic obligation to choose "good colors" for planets that maximizes their interesting features. We already did it for surface-views on Earth.

The words that sprung to mind for me were "valuable scapegoat", and hence a sci-fi quote:

Leo realized suddenly that Van Atta was not going to let him get a word of defense in edgewise if he could possibly help it. Van Atta was onto something infinitely more valuable than mere vengeance for a punch in the jaw—a scapegoat. A perfect scapegoat, upon whom he could pin every glitch in the Project for the past two months—or longer, depending on his ingenuity—and sacrifice qualmlessly to the company gods, himself emerging squeaky-clean and sinless.

-- Falling Free by Lois McMaster Bujold

Some fault lies with the users, therefore, the company is obviously utterly blameless. /s

Yes. The deliverable is not the work done. Imagine you a patch submitted to your code base, you have to decide whether to start deeply reviewing it.

You are given an envelope, with a message inside revealing whether the patch came from a human contributor with expertise in the field, or whether it was prompted with "make stuff better" out of an LLM.

Do you throw the "useless" envelope away without opening it? Authorship indirectly tells us things about other work that happened, like checking alternative approaches, checking for edge-cases, and considering future work.

Shout-out to javascript disabled by default :)

Trump, the first president with enough scandals to cause choice paralysis.

I wish i could just opt out of social security and invest the money myself.

I'd like to emphasize that SS/OASDI is not an investment program in the first place, it is an insurance program. The two kinds cannot be directly compared, and have very different mechanics and features. You might already have known that, but the misconception is distressingly common in America--which I blame on misinformation from big-bank lobbyists.

You inherit all risk. But you opt out of the taxes. You must prove retirements assets are being contributed to.

Is inheriting all the risk even possible? Suppose someone signs some dark legal contract in blood, like: "I refuse all public assistance, let me die in a ditch because I shall make my own fate."

What happens if they commit crime (perhaps out of desperation) and are sentenced to time in jail? Now there are three outcomes which all suck differently:

1. The government caves and supports them anyway with taxpayers funding their jail-food and jail-shelter. This distorts the original incentives, and recidivism is going to be a bitch.

2. They are let go, to commit more crimes? Locals won't stand for that.

3. They are indirectly executed by being forcibly exiled to a walled-off isolated place with no food and no shelter, as their families (quite reasonably) cry about the brutality on TV.

Then there's the issue of dependents, and mechanisms for fraud, and both of those are much bigger cans of worms than I want to open in this edit...

It sounds like a defined-benefits fund

The first key is that it isn't defined-benefits, in the sense that there is no account labeled "Bob Smith's Accumulated Retirement." Don't be disheartened though, because many Americans have the wrong idea too. [0 - See rant in footnote.]

In addition, the US has no constitutional barrier to protect it, the federal legislature can pass a regular law which completely rewrites the benefits however they like. It could be very unpopular, though.

Can someone explain the legal structures [...]

It has grown a lot of bells and whistles over time, but at the core its formal original name of "Old-Age, Survivors, and Disability Insurance" is very informative.

* The premiums for coverage are collected as a tax on the working.

* Payout conditions broadly involve being alive and not able to earn enough to stay that way.

* If you pay in and then die young and healthy, you don't get anything. This is normal and intended, the same way that home fire insurance doesn't pay if your house is swept away by a tsunami.

* The program's surplus funds (from planned-for demographic shifts) is invested in bonds with the US government, meaning that there's an intra-governmental credit/debt going on, where OASDI/SS is the creditor and government-in-general is the debtor.

[...] that make Social Security "run out"?

Most of the "run out" talk refers to a period of time where the invested surplus dwindles due to yet-more demographic shift, and cannot cover the difference between inflow and outflow. At that point one or both of these will have to happen:

(A) Congress passes a law increasing premiums/taxes on current workers

(B) Congress passes a law saying it's OK to pay less than the program did before.

Congress has been procrastinating on this for many decades.

__________

[0] I blame this on deliberate tactics by big-banks, and political groups ideologically opposed to the program. Private banks are unable to make big bucks offering a competing insurance plan, so instead they promote a false comparison. It goes like this:

1. They falsely assert that X% of the current surplus is somehow already exclusively "yours."

2. They claim that "your" money exists in a boring lame government retirement account which only invests in bonds. (Only half-true, in that the surplus is in bonds.)

3. They ask if you'd rather have the option of moving the money to a new account run by Big Bank, who is so much cooler will help you (for a modest fee) invest in stocks which go up much faster so "your" money will be zillions by the time you retire.

Huh, if that site is indeed theirs, that means they are now accusing strangers (me) of fabricating quotes from their own pitch-deck!

That raises some interesting questions about the creation process. I like to believe that authors should recognize their own works.