Problem is that the frontier models are nowhere near 99% reliable. Orchestration and good system design is how you get reliability. Yes, the frontier models still are going to be better by default than open source models. But the LLM is still only a component in a broader system. What's seeming to be actually necessary for any high-usage worthwhile use case is making your model task specific (via fine-tuning / post-training / RL). I build these systems for enterprises. The frontier models are not enough.

> The whole concept of buying services from people is either that their time is worth less than yours

And if everyone else's time has become more valuable then too has the time that is being saved by buying services.

If my time as a programmer is worth significantly more now than it was 25 years ago, then the time I save by buying services is worth more.

There's a reason that someone making $1mil/year is going to be willing to pay more for the exact same haircut that someone making $70k/year also gets. The time being saved is worth more to them.

You're only looking at half of the equation here. Following your logic, if my time is worth $100/hr, I should be willing to pay $99/hr for a haircut. But reality is that a haircut isn't just worth some utility value based on time saved, it's worth the lowest amount where suppliers' willingness to provide it at a given quality and buyers' willingness to pay meet.

So while the $99/hr haircut might technically save me money/time, suppliers of haircuts are generally willing to give the same haircut for $30/hr. If one supplier tried to pin their prices to the growth of their customers' income, they would go out of business. That is because the value of the suppliers time isn't increasing at the same rate.

I'm mostly bald... I got tired of paying as much as I did for haircuts and now mostly just use a pair of clippers on myself, since my goal is to take off all of it. I've paid for more beard trims the past few years than haircuts, though I mostly do that myself too.

Note, I usually use clippers on myself about once a week. Sometimes I'll use a shaver to get a closer shave, but generally doesn't matter as I don't care if there's a little growth, which is noticeable unless I literally shave daily anyway... which I'm too lazy to do, and definitely not able to pay someone else to do.

Yea I've been seeing very similar behavior from people. They think of themselves as static, unchanging, uncreative but view LLMs as some kind of unrelenting and inevitable innovative force...

I think it's people's anxieties and fears about the uncertainty about the value of their own cognitive labor demoralizing them and making them doubt their own self-efficacy. Which I think is an understandable reaction in the face of trillion dollar companies frothing at the mouth to replace you with pale imitations.

Best name I could think of calling this narrative / myth is people believing in "effortless AI": https://www.insidevoice.ai/p/effortless-ai

Very well said. I think it's hard for people to understand that ADHD is simultaneously over diagnosed and under diagnosed.

This does make sense given modern understandings of ADHD are that it's not about it being a deficit in attention but rather a deficit in self-regulation and executive functioning.

I focus, and hyperfocus, without any help from a stimulant.

In fact, I am far more likely to go deep into a rabbit hole, of useful or wasted time, when I am tired.

But with a stimulant it is far easier at any time to choose to focus on ordinary important tasks, follow through, and enjoy getting necessary things done. I.e. improved executive function.

> For transparency, a single software engineer budget is $670K+.

Are you saying that the costs to employ a single software engineer is $670K+? If you mean something else then nvm.

Otherwise that's a ridiculous number to use unless you are specifically talking about places with the highest cost of living in the country where a mid-level dev starts at over $200K.

I am saying that. Salary + taxes + insurance + retirement + other benefits + support cost is around 670k. Salary eats up like 160k of that budget, though.

Please provide a breakdown.

I find this 4X of base salary implausibly high. 2X strikes me as closer to my reality at a large academic medical center.

I don't have a breakdown. It was a number cited to me from a manager. Downvotes are interesting.

Payroll taxes on $160k salary are $12,240. Employer contribution to health insurance is maybe $6k - $20k. Retirement maybe $5k. Still under $200k.

Heck of a lot of "support cost" to get to $670k

Your numbers make sense from what I've seen in private sector. And meet the common sense threshold as well.

Whether the numbers are either wrong or if that is truly what support costs look like at a university would be interesting to know.

> My concern is that, even if the exercise is only an amusing curiosity, many people will take the results more seriously than they should, and be inspired to apply the same methods to products and initiatives that adversely affect people's lives in real ways.

That will most definitely happen. We already have known for awhile that algorithmic methods have been applied "to products and initiatives that adversely affect people's lives in real ways", for awhile: https://www.scientificamerican.com/blog/roots-of-unity/revie...

I guess the question is if LLMs for some reason will reinvigorate public sentiment / pressure for governing bodies to sincerely take up the ongoing responsibility of trying to lessen the unique harms that can be amplified by reckless implementation of algorithms.

The RES (Reddit Enhancement Suite) browser extension indirectly does this for me since it tracks the lifetime number of upvotes I give other users. So when I stumble upon a thread with a user with like +40 I know "This is someone whom I've repeatedly found to have good takes" (depending on the context).

It's subjective of course but at least it's transparently so.

I just think it's neat that it's kinda sorta a loose proxy for what you're talking about but done in arguably the simplest way possible.

I am not a Redditor, but RES sounds like it would increase the ‘echo-chamber’ effect, rather than improving one’s understanding of contributors’ calibration.

Echo chamber of rational, thoughtful and truthful speakers is what I’m looking for in Internet forums.

That’s what everyone living in an echo chamber (and especially one of their own creation) thinks they’re in.

"you're in an echo chamber" is one of the most frightfully overused opinions.

The expression is an echo chamber in and of itself; it is self-fulfilling prophecy.

I don't think I'm in any is my problem (HN is better than most, doesn't mean it's good in absolute terms...)

flat earth creationists would describe their colleagues the same way.

a group of them certainly is an echo chamber; why isn't your view?

He doesn't deny that his point of view forms an echo chamber.

"flat earth creationists would describe their colleagues the same way."

Actually they mostly don't. Lots of infighting over the real true answer .. (infinite flat earth, finite but with impassable ice walls, ..)

An echo chamber is a product of your own creation. If you're willing to upvote people who disagree with your and actively seek out opposite takes to be genuinely curious about, then you're probably not in an echo chamber.

The tools for controlling your feed are reducing on social media like Instagram, TikTok, Youtube, etc., but simply saying that you follow and respect the opinions of a select group doesn't necessarily mean you're forming an echo chamber.

This is different from something like flat earth/other conspiracy theories where when confronted with opposite evidence, they aren't likely to engage with it in good faith.

it depends on if you vote based on the quality of contribution to the discussion or based on how much you agree/disagree.

I don't think you can change user behavior like this.

You can give them a "venting sink" though. Instead of having a downvote button that just downvotes, have it pop up a little menu asking for a downvote reason, with "spam" and "disagree" as options. You could then weigh downvotes by which option was selected, along with an algorithm to discover "user honesty" based on whether their downvotes correlate with others or just with the people on their end of the political spectrum, a la Birdwatch.

You can't change it for other users, only for yourself, which is what the original comment about the extension said.

Echo chambers will always result on social media. I don't think you can come up with a format that will not result in consolidated blocs.

Reddit's current structure very much produces an echo chamber with only one main prevailing view. If everyone used an extension like this I would expect it to increase overall diversity of opinion on the site, as things that conflict with the main echo chamber view could still thrive in their own communities rather than getting downvoted with the actual spam.

Hacker News structure is identical though. Topics invite different demographics and downvotes suppress unpopular opinions. The front page shows most up voted stories. It's the same system.

HN's moderation and ranking is better. But there's definitely an echo chamber effect here too.

HN has some built-in ways to reduce this, like not allowing everyone to downvote everything.

More than having exact same system but with any random reader voting ? I'd say as long as you don't do "I disagree therefore I downvote" it would probably be more accurate than having essentially same voting system driven by randoms like reddit/HN already does

That assumes your upvotes in the past were a good proxy for being correct today. You could have both been wrong.

To be clear...prior to this recent explosive interest in LLMs, this was already true. Snowden was over 10 years ago.

We can't start clutching our pearls now as if programmatic mass surveillance hasn't been running on all cylinders for over 20 years.

Don't get me wrong, we should absolutely care about this, everyone should. I'm just saying any vague gestures at imminent privacy-doom thanks to LLMs is liable to be doing some big favors of inadvertently sanitizing the history of prior (and still) egregious privacy offenders.

I'm just suggesting more "Yes and" and less "pearl clutching" is all.

Who, exactly, is the "we" who you see "pearl clutching" instead of "yes and-ing"?

I personally don't think belief in an afterlife should be necessary to believe it's worthwhile to not be shitty to people.

"What goes around comes around" suffices for me.

Call it "ethics", call it "maximizing outcomes for all involved stakeholders", call it "karma", "good business", or "kindness"...whatever you call it, I don't think it's difficult to find your own personal justification for it if you want to.

> "What goes around comes around" suffices for me.

As you say. Best wishes, in any case.