ozr

I'm sorry this is hard for you to follow. Good luck elsewhere, maybe someone else will be able to help you understand.

I agree. High performers should be recognized via whatever process the company has to do so. Leveling, at reviews, etc. The lack of doing that is just shitty management. It doesn't change the basic description of the market above.

This is true. It's a result of poor management, as you said. Either a lack of incentives or a lack of skill.

If someone leaving represents a substantial risk, then a) that role should be derisked, and b) that individual should be recognized via whatever process the company has (leveling, etc) to keep them on board.

Is it inequitable and mistreating an employee to pay them what we agreed to?

> I don't think what you're doing provides a "why", really

I disagree.

An employer has a contract (whether literal or not) with an employee to pay a certain amount. They continue to do that, along with nominal raises. This is reasonable.

If they want to hire someone new, that new person will generally demand the market rate, since they can get it elsewhere. This is also reasonable.

The only thing that bothers people is that both of these things occur at the same time.

> I'm not sure I quite understand what you mean by value here, exactly. Value to the employer? That doesn't track the CPI. I think you must mean value to the employee

I mean the value to the employee. Maybe getting slightly too wonky here, but to elaborate: if I hire you, I may offer you a certain amount of US dollars per year. You don't actually care about the dollars. You can't eat them, you can't live inside the bits of paper. You and I only care about them as a store of value that can be conveniently exchanged: a currency. What I'm actually paying you is some amount of value each year.

This is imperfect as a result of inflation. If I pay you the same amount of dollars per year, the value that I'm giving you goes down. So there is some sort of annual raise, generally referred to as a cost of living adjustment, that is supposed to account for that.

In the past couple of years, it hasn't at a lot of places. This is effectively a pay decrease, and is very uncool. Historically, a 2-3% CoL raise per year _would_ match inflation.

I'm pretty sure you know all that, but it leads into:

> Think about it in the market for apples rather than the market for labor, for example. Obviously it does work different in the labor market, but you have to say what the relevant differences between the market for apples and for labor are to explain things.

I don't think I can give a meaningful analogy here. Aside from the shelf life issue, participants in the labor market have agency, whereas an apple doesn't. But, in an attempt to work through it:

If I want to go to the labor or super market to get a new employee or apple, I have to pay the price they are asking. If I don't, I don't get it.

If I already have an employee or an apple, I don't need to pay the market rate. I already have it. I am no longer participating in the market, so the prices there are irrelevant to me.

This is where the major difference manifests: an apple in my cupboard is not going to leave because someone else will pay more for it. An employee might. Simply because I'm not in the market doesn't mean they aren't.

The odds of them leaving, compared to the cost of responding to market fluctuations in their favor only, is generally going to be net negative. If it was not net negative, this wouldn't be a topic of discussion.

In the event that I _know_ it would be painful for a particular employee to leave, most companies have a process for recognizing that and giving an outsized raise, one that meets or exceeds the current market. 'Retention risk', 'Exceeds Goals', giving a promotion or level increase, etc.

All that being said: this is strictly from an econ point of view. I very much don't like thinking of my reports this way, and I don't do it in practice. But when I'm talking with HR and upper execs, this is how they understand it.

> Should why, though? You didn't explain why they should have this expectation, rather than any other.

Because if we agree that I pay you $X/yr, that's just a placeholder for a certain amount of value per year. As inflation rises, the nominal salary I pay has to increase to maintain that value. A CoL raise handles that.

> There is a pressure to keep these in lock-step: if you are paying ...

This is all theoretically true, but just not how it works in practice. See: this post and all the comments lamenting it.

I always figured the explanation was much simpler than this.

As a general thing, someone meeting expectations and performing their current role should expect a nominal raise each year. Cost-of-living sort of thing.

When you hire someone new from the market, you need to figure out the market price and pay that.

There's no particular reason these two should be in lockstep besides, as the author mentions, a vague sense of inequity.

The process of recognizing that a subset of a component has failed, recovering from the failure, and resuming would result in a slow-down. I'm not sure that's a full explanation for the experience though.

Excluding people 16-24 would definitely remove the vast majority of the population addressed above.

This is cute, but not data. When you look at actual data, it shows that it actually has changed over time. There is a gradual rise in the labor force participation rate as more women enter the workforce, up until a peak in ~2001. It's been dropping ever since.

https://fred.stlouisfed.org/series/CIVPART

Depending on your choice of measure, they are either greatly beyond their parents or mildly behind.

More importantly, why is that a reasonable expectation?

And, as we know, there are no cultural changes between generations.

They've got the vast majority of their business in 'traditional' quads, it's their core competency. This is a bit of a lark which could be very cool, but making the investment in a new platform that's very different specifically for it is probably not worth the cost and complexity.

A pathological sort of comparison is much more common now. They assume that what their parents have (with 20+ years of work), or what they see on social media, is what they should have too. That they don't is perceived as a great injustice.

You can see someone that has what you want and either a) work towards it, or b) complain.

There's always a mix of both, but b) is very loud right now.

Even beyond entertainment, interacting with things that are aesthetically pleasing is fundamentally good for individuals and society.

It won't help as much with local models, but you could add an 'aligned AI' captcha that requires someone to type a slur or swear word. Modern problems/modern solutions.

The term is spreading, unfortunately.

That's reading a lot into nomads that go to low COL countries. There are plenty of reasons to do it beyond an artificial crown and 'sexual privilege'.

You're able to take jobs that you may enjoy more for less pay. You can keep your personal burn rate lower while you're working on a startup. You might even just like seeing different parts of the world.

This is just wrong. Looking at 2016, the US isn't even in the top 10 per capita, and it's behind 4 EU states.

https://www.worldometers.info/co2-emissions/co2-emissions-pe...

Why would you do that, and when else has it worked?

I really like Grammarly as a product, but I exclusively use their web editor. I wonder what their web vs desktop usage is.

Not really with our current techniques. The increased latency and low bandwidth between nodes makes it absurdly slow.

Those are all ultimately represented as bytes. You're just looking for a different query language.

We have this right now. It's abstractions on top of the real primitives. That's what client protocols and query languages are.

Ah, ok, I read that wrong. You're right. Thank you.

I'm not sure how this connects. Diffusion models are great for certain tasks, not for others. They have sort of a niche use in audio and are the SOTA for image generation, but there is a lot of education (almost all) that doesn't apply to those domains.

When we're talking about multimodal ML systems, the core is generally a transformer-based LLM with no diffusion component.

This simply isn't true. Looking at the IMF numbers in 2023, the US is number 7. Every country ahead of the US has something that is massively distorting their GDP per capita. Tax havens, nationalized oil, etc.

The police are interested in doing their jobs. They are not allowed to by a local government that will harangue them for trying, and district attorneys that won't prosecute or push for meaningful jail time.

Madoff got 150.

I don't really like voice interfaces, but I see that as a personal preference.

What sort of limitations/restrictions do you see with it, beyond accuracy (which is continually improving)?