Monday, June 2, 2025

LLM's: the backhoe and the bench-press

This is in response to a post at Brad DeLong's "Grasping Reality" blog, From ChatGPT back to clay & cuneiform: A start at rethinking pedagogy for the Age of "AI".

https://wiki.opensourceecology.org/wiki/File:Backhoe.png

He asks himself two questions about broad objectives as a teacher:

  • What do I want my students to remember five years from now about this class?
  • What do I want my students to know how to do after they have taken this class?
Those seem like absolutely the right questions.

In thinking concretely about his course in economic history, he presents his week-by-week themes and wonders about a weekly structure of three parts:
  1. What is our question?
  2. How would you go about discovering the best answer to it, at least to your satisfaction?
  3. How would you then persuade somebody else that the answer you have come up with is the most likely answer?
He writes,
Could that serve as a scaffolding for education in the age of AI?

I think it could. By focusing on 'how would you discover?' and then 'how would you persuade?', we could avoid all of the traps involved in the overuse of LLMs.
This seems promising to me, though I suppose the first test would be to take items 2 and 3 and ask an LLM to do them.

My broader concerns about the impact of AIs on learning are rooted in how I view the role of memorization and practice. And the difference between short-term and long-term memory.

Monday, May 19, 2025

What's in a record

Wandering around the Web, I found an interesting observation in a discussion of climate change and the reluctance to accept the scientific consensus.

A commenter wrote:

The record high for India was in 2016. Check out these record highs from around the world. Far more highs a few to several years ago than within the past couple of years. What does that tell a thinking person?

And they gave a link to a Wikipedia page with world weather records: https://en.wikipedia.org/wiki/List_of_weather_records

Which got me wondering, What does that tell a thinking person?

I scrolled through the list of record highs and noticed that, indeed, there were a lot of dates from before the 21st century and thought that was interesting.

Then I started wondering what we might expect the data to look like if there were no warming.

And my thought was that they should be evenly distributed through time.

So I threw them in Excel and took a look:

And I thought, "That doesn't look very random!"

Sunday, April 20, 2025

After Social Security: The elegant balance

The previous post—the first post in what might be a series—gave an overview of what I think is important to understand about Social Security, and then dug down into the logic of why every retirement system is, of necessity, a way of moving claims on current output from people currently working to people who are retired.

This installment looks at how Social Security elegantly balances competing visions of what it means for this type of system to be “fair.”

Why do you earn what you earn?

Think about why you have the income you have, compared to someone else’s.

Maybe you get paid $40,000 a year. Or (less likely), it’s $200,000 a year.

Why do you have one and not the other?

With some pretty strong simplification, and maybe a bit of arbitrary line drawing, we can put the causes into two buckets.

The first is merit: your hard work, your intelligent decision-making, your creative problem-solving.

The second is fortune: good luck lifting some people up, bad luck dragging others down.

Remember that Social Security runs off of a “flat” tax, where everyone pays the same 12.4% of their payroll income, up to $168,000 of income (so maxing out their Social Security taxes at $20,832).

Given that people are mostly putting in the same portion of their income, how much should each person get back?

That “should” depends on your sense of the mix of merit and fortune.

Tuesday, April 15, 2025

After Social Security

Original image from By Social Security Administration - Social Security Administration, Public Domain, https://commons.wikimedia.org/w/index.php?curid=43304462
Modified by the author


Social Security isn’t necessarily dead, but there have been Republicans trying in various ways to end since at least 2005, when George W. Bush came out of his re-election saying that he had won political capital and he intended to spend it.

Of course, that effort ended up going nowhere – Bush spent some political capital in his failed push, and then burned through a bunch of what was left with his botched response to Katrina later that year.

Now we’re back at it and the threat is arguably sharper. We’re dealing with a Republican party that is less independent from its president than 20 years ago, a president less inclined to listen to reason, and an off-brand Tony Stark who’s been given the keys to the kingdom and may well have the capability to make the Social Security system so dysfunctional that it falls apart without Republicans in Congress needing to leave their fingerprints on the murder weapon.

Social Security’s demise is hardly a foregone conclusion, but it’s enough of a possibility that I thought it was worth looking at what our options might be after it happens.

The part of this post above the fold is the TL;DR of the whole thing, a series of declarative statements.

Below the post is the “too long” explanation of that first declarative statement.

I hope I’ll lay out explanations of the rest over the next few weeks, but there’s usually a large gap between my intentions and my accomplishments.

Laying out the spine of the argument is easier, so here it is.

There are lots of different ways of organizing a retirement system: government pay-as-you-go (PAYGO) like Social Security; private PAYGO like an old-fashioned company pension; pre-funded company pensions, and private retirement accounts. But the differences among those systems are differences in accounting. Underneath, every single retirement system has to do the same thing. It has to move claims on current output from people who earned those claims by working to produce current output, to people who are no longer working and therefore not playing a role in the production of current output.

A retirement system might want to give a helping hand to people who, for whatever reason, had relatively low lifetime earnings. It also doesn’t want to incentivize people not to take care of themselves during their working lives. Those two goals are potentially in conflict, but Social Security strikes an intelligent balance between them.

People describe the Social Security trust fund as a fiction or a Ponzi scheme, the government borrowing money from itself to pay itself. That’s sort of true, but it’s more false. The 1983 reforms that led to the large (but shrinking) trust fund we have today represented an implicit bargain about who on the income scale would pay when. People of average or lower income would pay up front in the 1980s, 1990s, and 2000s, in order to build up the trust fund, so that Social Security was there for them when they retired. People of higher incomes would pay in the 2010s and after in order to fulfill that promise. Destroying the system now is letting higher-income folks out of their end of the deal, just when it becomes time to pay. And the part about it being a Ponzi scheme—“There’s no money in there for you when you retire! The money you pay now just goes right at the door to pay today’s retirees!”—that part is simply based on a failure to understand what a retirement system does, which is that it moves claims on current output from people who are working, to people who are no longer working.

Replacing Social Security with individual accounts has three problems. First, individual accounts have administrative costs that are an order of magnitude higher than Social Security. Second, the intelligent balance between cushion and incentives described above is impossible with individual accounts. And third, the promise of higher returns by putting money into the private market instead of Social Security is logically impossible to achieve on a system-wide basis, because of the unavoidable nature of every retirement system (moving claims on current output from people working to people who are no longer working).

A large part of why Social Security has financial problems is that more and more of the claims on current output are going to people in forms that don’t get taxed to cover Social Security. That’s a fixable problem.

Social Security faces three fundamental problems. First, an increasing share of the population is of retirement age. Second, the growth of GDP per capita is slowing down. Third, the system as currently funded will run out of money. That third one, as mentioned above, is fixable. The first two don’t have such obvious solutions, but they are just as serious for any alternative to Social Security, because the underlying nature of any retirement system is that it moves claims on current output from people who are working, to people who are no longer working.

Lastly—and to justify the title of “After Social Security”—if Musk et al. succeed in destroying Social Security, it will be disastrous, but it doesn’t have to mean the end of sensible retirement arrangements forever. The bad part is, they will have stolen a lot of money and impoverished millions and probably led to a significant amount of premature death as people face challenges they could have overcome had they had access to the Social Security benefits that were their due. But any retirement system is ultimately a social arrangement, one that (say it with me now) moves claims on current output from people who are working, to people who are no longer working. The true wealth of a country is not measured in collections of financial assets, but in the ability to accomplish useful things. To the extent that we still have that ability after Musk is done experimenting on us, it will be in our hands to decide whether to re-establish a decent retirement system.

That’s the spine of the argument, and it only took me about 750 words.

The rest of this post will be making good on the first claim, the thing that is most important of all to understand before trying to solve Social Security’s problems: Every retirement system moves claims on current output from people who are working, to people who are no longer working.

Sunday, March 23, 2025

Apocalypse no. 2

 Part of a semi-regular series. The first part is here.

Albrecht Dürer
The Four Horsemen, from The Apocalypse
via Wikimedia

You've probably encountered people saying we don't need vaccines because "who gets measles now anyway?" and "when was the last time smallpox was an issue?".

You get why that's a problem, right?

We don't see measles moving through our population because almost everyone has been vaccinated against it, and we're in the middle of getting to see what happens when you drop that rate from 95% to 80%.

This is the paradox of vaccine success. They work so well, we forget that they do anything at all.

What if we had this attitude towards the way society as a whole works?

Government is a convenient whipping boy: a target for our frustrations with how our lives are going. And it's easy to find things that government does wrong, could do better, or makes worse.

Unfortunately, it's even easier to overlook the basic things that government does well which enable the day to day lives we've come to expect.

For a quick and morbid example, it's illegal to sell baby cribs with bars too far apart. If you're a manufacturer of baby cribs, you might find this annoying. If you're a parent of an infant, you might appreciate that the government has made it impossible for your baby to get its head stuck between the bars of its crib and strangle itself. But you also might be unaware that such a rule exists. All babies - including yours - are now safe from getting their heads stuck between crib bars and strangling themselves to death, and people don't realize that it was government action that removed that risk.

For a longer example, and one that hits home given where I live, we need an outline understanding of how a rural economy works.

The one economic field in which rural areas have a natural advantage is resource extraction.

Friday, March 21, 2025

Apocalypse no. 1

 Part of an occasional series

Albert Goodwin, Apocalypse
https://en.m.wikipedia.org/wiki/File:Apocalypse-Albert_Goodwin.jpg

You could say I'm doomer adjacent.

I have friends, acquaintances, and colleagues who are full-on doomer, and I understand their arguments.

On one side, there's what we're doing to the natural systems that support us. At the broadest level, global warming of course, but we're pushing at ecosystems and resources in lots of other ways as well. This is a problem of having too much access to fossil fuels for our own good.

On the other side, the idea of peak oil raises the possibility that we don't have enough easy access to fossil fuel to maintain the way of life we've come to take for granted.

Too much and not enough at the same time. I understand the argument, and I can't prove it's wrong.

I used to be full doomer myself, and in 2008-09, I expected that's what we were seeing. I was clearly wrong, so I've learned to be more cautious about forecasts.

Understanding the deep forces beneath a situation is useful, but does little to help you predict the details of how things will unfold.