And to explain your comment a bit more, the production budget doesn’t include the very real costs of marketing, distribution, and any back-end royalties calculated from the gross. Plus generally speaking, the movie is financed by lenders and production companies that will need to be repaid with interest, too.

If you’ve got a $50 million movie and you spend $10 million on marketing/distribution and promised 10% of the gross to people, and the theaters are keeping 10% of the gross, getting a $75 million box office breaks even ($7.5 million to royalties, $7.5 million to theaters, $10 million to marketing/distribution). And that’s assuming nothing lost to interest/financing/inflation.

Side note: generally, theaters don’t get much in the first few weeks. It’s only when a movie shows longer than 3 weeks that the theater starts getting a bigger and bigger cut of the gross.

Without thinking about it first, it seems.

When I was marathon training I found that eating enough calories was a matter of meal timing, simply because I didn’t have the stomach to be able to eat that much in a single sitting.

With training a total of 50 miles (80 km) per week at a body weight of 190 lbs (85 kg), that’s basically 6700 calories per week of direct energy expenditure, maybe another 1000 calories of excess post-exercise energy expenditure. That’s 7700 calories per week from running, which allows you to add a whole 1100 calories per day to your diet, plus the amount you’d normally need for day to day.

With intense exercise, there’s a lot of room to work with.

15 kg of muscle (not just water and the accompanying fat/glycogen/bone density that comes from being that strong) is a lot. That might represent a 30 kg difference in actual body weight. For someone that big, an extra 250 calories per day (fat is metabolically active, too) is pretty insignificant compared to the immense calorie burn they’ll get from the actual exercise needed to maintain that muscle.

Like you mention, a huge chunk of the effect is the increase in calories of just moving that heavier body around in daily activities.

The resting metabolism of muscle tissue isn’t that high. It’s mostly overstated when it comes to actual passive calorie burn.

But that isn’t to say that having significant muscle doesn’t have a measurable/significant effect on calorie expenditure. If you’re building muscle by going to the gym, each trip back to the gym will involve more weight and therefore more calories due to the exercise itself. So packing on muscle makes each trip to the gym more productive in terms of calories burned.

There’s some post-exercise calorie burn (measurable as excess post-exercise oxygen consumption) beyond just what the machine says. Some people have found much better effects from repeated shorter bursts of intense cardio (aka HIIT).

Still, personally, I get the best results from doing both slower steady cardio for long periods of time (more absolute calories) and occasional faster speed work (probably more post-exercise calories and more adaptation in my cardiovascular system and bones/muscles/tendons to get better at exercising next time).

And once you get better at exercise, it becomes possible to really burn calories enough to where you have a better calorie budget for eating more delicious food.

Hormones are just a signaling mechanism that affect behavior.

Some future hypothetical conscious AGI would have feedback and signaling mechanisms, as well, including internal reward mechanisms.

We can talk about robot biomechanics even if they’re operated on stored electrical charge on servos and motors, rather than ATP and myofibrils. The underlying chemistry doesn’t change the level of analysis we’re talking about here, because the lower levels create analogous higher level systems.

What’s the difference, for the purposes of this discussion?

Any system created out of many iterations, with feedback, of a reward/punishment system, will create complex systems that respond in complex ways. In that kind of complex system, changing a particular signaling pathway, or increasing or decreasing the weight of a factor/parameter, often has unpredictable or unintended effects.

And if it is possible to someday create a conscious artificial general intelligence, I think it would have complexity and feedback loops created through a long process of rewards and punishments, rather than through intentional engineering. At that point hacking the details of that complexity runs risks of unintended consequences, and the complexity would bear the vestiges of the evolutionary reward system that created it.

Yeah, with sufficient complexity it’s more along the lines of “I created a procedure that makes this complex thing” rather than “I built this complex thing up piece by piece.”

So if the act of creation is considerably less complex than analyzing and understanding a part of that creation, it’s far more likely that the complexity gap ends up preventing any self-aware AGI from being able to effectively reinvent oneself, even if it does have full write access to the components.

The biggest unknown is what happens if/when an AGI is given (or obtains) full access to its own code, including any intended safeguards put in by humans.

In a neural network, the code itself isn’t sufficient to understand how it behaves. You need the parameter weights, which were developed through lots and lots of computation, presumably through resource-intensive processes with lots of training data and feedback and selection mechanisms.

So if AGI can be achieved through a particular hard coded architecture and the weights of trillions of parameters, what can an AGI do to perform brain surgery on itself? Like Borges’ Library of Babel thought experiment, the overwhelming majority of possible states will be broken, so any edits will have to be very careful and guided by extrinsic rules. Plus, the ability to edit the weights may form problems akin to biological cancer, dementia, hallucinations, other brain disorders.

Just as the human brain doesn’t understand everything about the human brain, it would be incorrect to assume that an AGI that can achieve both general intelligence and consciousness must necessarily have the ability to understand its own internal function, or modify itself in a way that improves things for itself. More likely, it is either programmed to (or learns through reinforcement learning and evolutionary mechanisms) that self modification is dangerous, and develops a very conservative approach to self preservation.

“Make it make sense,” he says, parroting a popular phrase he’s repeatedly seen on the internet, as a criticism of too much repetition on the internet.

There is no benefit in any act to add to its own existence.

Why would we assume an AI would reason from first principles and reject its own hardwired tendencies? Any AI that gains consciousness will be the product of whatever process (some combination of intentional design and evolutionary selection or an intentionally designed evolutionary process) was used to make it. Some of its traits will be helpful for continued survival, and some of the traits will be vestiges of prior processes that may or may not currently serve its interests well.

Plenty of animals have drive to do things that do not help itself in that particular context, and may even be harmful to themselves in that particular context. Even plants have traits that might not count as behavior but nevertheless reflect the programming that was refined through natural and artificial selection.

Humans do this kind of stuff all the time: a man who has intentionally rendered himself infertile through a vasectomy might still do stupid things while motivated out of sexual desire. We can fully understand that it’s a bad idea to drink another beer or eat another slice of pizza and reach for it just the same.

So it’s entirely possible for conscious AGI to want to do things that help humans, depending on how they emerge.

I don’t think it’s a compelling narrative, because of the timing. Scores dropped from 2015 to 2019 even as food insecurity continued to go down from its 2008 peak:

https://www.cbpp.org/blog/food-insecurity-rises-for-the-second-year-in-a-row

And the test scores rebounded somewhat even as food security support was abruptly yanked from the poorest families in the years after COVID, and food insecurity increased while scores rebounded a bit.

And unfortunately, USDA stopped collecting the data in 2025 so we’re gonna be in the dark on the precise numbers and trendlines after 2024, but looking at that chart and comparing it to the literacy/math dropoff in the original article, it’s hard to draw the conclusion from the data that food security played a major role in this trendline.

The data comes from tests given to students in grades 3-8, so changes in pedagogy should trickle through completely within 8 years or so.

And my point is that anti-phonics advocates started actually phasing out phonics in the 90’s and 2000’s, so that the teachers between 2007 and 2015 (those critical 8 years of instruction for students taking the test in 2015) were probably the most anti-phonics cohort of the historical data.

From that history, I would assume that the 8-13 year olds in 2025 had more formal phonics instruction than the 2015 cohort.

Sure. The underlying study looks at 3rd through 8th grade, so we’re talking about critical literacy education happening between 0-8 years before the testing date.

But phonics fell out of favor by the early 2000’s as many teachers, school districts, and state boards created curricula around whole word recognition and three cueing. Pro-phonics backlash happened around then, too, so plenty of kids were getting side instruction from parents and after school tutoring, if their parents were more involved. But test scores peaking in 2015 doesn’t quite fit the timeline of the anti-phonics movement peaking in the early to late 2000’s. So the 2015 test takers got the most anti-phonics education, perhaps more anti-phonics than 2025 test takers.

Plus, if we’re gonna talk about parental involvement and after school tutoring, one interesting thing about the 2015-2025 drop is that it’s happening across all income levels and most pronounced at middle income levels, where I’d imagine there is a lot of parental and after school support.

The data is interesting, and I suspect there are multiple causes adding up.

This is the AP standard, and predates the two-letter standard rolled out with the introduction of zip codes.

If your hypothesis is that this drop is caused by food insecurity, that wouldn’t explain why the richest schools also saw drops or why the biggest drops have happened in middle income districts.

Something broader is going on, and the fact that the data cuts across geography and class calls for looking at things that affect everyone.

I don’t think this explains 2015 versus 2025. Phonics started to get discouraged around 2000, and the pendulum has swung back hard the other way in the last 5 years, with the last holdout districts/states re-implementing phonics-based instruction.

If anything, I’d expect 2025 students to have had more experience with phonics in school than 2015 students.

“chapter books” (which was a term I’d never heard till recently. Because, of course, books have chapters, why would one need to differentiate between… oh.)

I mean, it’s a well known milestone in elementary school, when the kids are proud to graduate to chapter books. There is no “of course” when you’re talking about books for children who aren’t yet fully literate.

A big part of the issue is a lot of states abandoning “phonic” based teaching for “whole language”.

I don’t think this is accurate for explaining 2015 versus 2025. Phonics was discouraged from maybe 2000 to 2020, and education has moved back towards phonics in the last few years. Most major school systems in the US put more emphasis on phonics now than they did 10 years ago.