Intelligence Decay

[Riggerjack]

And to riff on what campitor said...

What we do, we improve. However one solves problems, that path to a solution is reinforced. Next time a similar situation is presented, a similar path will be tried. The faster one learns, the faster and more automatic this process is.

For me, this felt like I already knew the best way to do anything familiar, and rarely rethought processes. I tended to skip from problem directly to solution, without processing all the decisions between problem and solution.

A few years ago, I noticed that these bright efficient paths weren't always ideal. That some groupings of problems may not be ideal. That I was forcing the solutions that worked so well for similar problems, and rounding errors.

So I have spent the last year or two trying to break open those bright pathways, separate out the ideal problem/solution combinations, and breaking the threads that were non optimal. This increases the likelihood of processing similar problems anew, rather than following a path that works, but not as well as it could.

Now this process feels stupid. I am intentionally making problem solving slower and more difficult. I feel a bit stupid. But it needs doing if I want better solutions.

The end result is that my brain was optimized for learning presented ideas in my youth. And I was very good at it. I thought it made me smart. (Evidence to support this theory is much harder to find when looking back that I assumed it would be... ) but as I then tried to apply what I "knew", in the world, well let's say there is a difference between theory and practice. Many of my bright pathways are where I grouped problems to match my education, my notions of how the world is and should be.

But just because I see the world in a particular way, does not make the world change to match my model.

What I am trying to explain is the difference in types of tasks and learning that happens later in life is very different from the tasks and types of learning we engage in as young adults, and completely different from learning presented information as youths. Skill in one circumstance often fails to translate to other circumstances. This isn't a drop in intelligence, this is development of new skills.

[Kriegsspiel]

I'm one of the people who feels smarter as they age.

ISo exercise the brain. Learn a new skill, do puzzles (sudoku, crosswords, etc.). Memorize new facts, read new books, learn to play an instrument. This will keep your brain healthy. Brain atrophy doesn't need to happen as you age. You're brain isn't slow because you're old, you're slow because you don't use the ole brain.

If you can get a hold of an old LSAT prep book, they're basically a giant brain teaser book.

[Kriegsspiel]

I also credit my unfolding intellectual supernova to regular binge drinking. I killed off my weak neurons years ago.

[Sclass]

A certain kind of my intelligence disappeared by the time I hit 25. I had been told as an undergrad physicist that great physicists do their best work before 30. I think this was the kind of smarts that just evaporated out my ears if I had any.

What little I may have had got squandered at high school level math contests where kids solved fiendishly tricky yet simple math problems designed to be solved in fifteen minutes with five clever lines on math...once the secret trick was unlocked. I suck at that stuff today.

However, I got more effective at solving real world problems with age. This masked the apparent intelligence decay. My hit rate went up. My youth was a mixture of lucky wins interspersed with disappointment. As I got smarter (wiser?) about how I attacked big problems I seemed to rack up more successes.

Also the importance of real world problems eclipsed any gains acquired by solving tricky fifteen minute math puzzles.

I may have been quick as a kid but I generated a lot of heat with my actions. I failed a lot too at big things in the real world. I really piled up a lot of junk that just didn’t work. Piles of waste. Piles I walked away from in frustration. Problems that I think would be easily finished off by the present me. I recall leaving a pile of scrap wood where I had promised to build a deck, leaving a hole in the ground that was supposed to be a swimming pool, leaving three cars on blocks that I just couldn’t make run during my 20s.

I don’t like to say with age comes wisdom. Because I know a lot of unsuccessful old people who keep making the same dumb mistakes. Somehow I adopted some behaviors like grittiness, focus, nit picking organization, long term vision, patient preparation and deliberate execution that got me more positive outcomes as time went on. I was a disaster up till about thirty. I looked like I was getting smarter because my hit rate was improving.

I recall at some point in my late thirties preferring to hunt for existing signal processing algorithms online or in books rather than designing my own from scratch. I also purchased some code for decoders I knew I should have been able to design myself. I think this was an indication of intelligence decline though I excused it as laziness at the time.

So you have this decaying IQ superimposed with this increasing success efficiency that seems to look like gradual improvement all while I’m probably following my mother’s footsteps into early onset dementia.

[jacob]

Information processing speed peaks earliest, around age 18 or 19.
Short-term memory is strongest at about age 25, before it begins to drop around age 35.
The ability to accurately identify others' emotions hits its peak during the 40s and 50s.
Vocabulary skills reach their height in the 60s and early 70s.
https://www.inc.com/laura-montini/when- ... think.html

Conventionally, this is split into fluid intelligence and crystallized intelligence. Fluid intelligence measures raw speed which I suppose we can translate into reaction time (especially reaction time that involves a decision) and short term memory. Tests like Raven is supposed to measure this. Crystallized intelligence is what you build using experience (deliberate practice). INTJs may think of the former as clock frequency and the latter as the operating system.

Many don't care to build their operating system or build it badly and crystallized intelligence does not necessarily grow with age at the same rate in different people. Think of it as a kind of 10,000 hours thing. Some waste the time and some become really good.

Overall, increasing levels of crystal should compensate for decreasing levels of fluid ideally flattening the performance if not increasing it slightly over the years. Consider solving a crossword puzzle. You need the raw fluid speed to find the word, but you also need the crystallized vocabulary in order to have something to find.

1 2 3 5 8 13 ?? measures something closer to fluid intelligence (although it does draw on your ability to add numbers in your head) than unscrambling "VIFRETU" into a word (especially in a non-native language) which draws on more crystallized intelligence to the degree that many adults might not even recognize the solution even after the word is pointed out to them.

After age 25 or so maintenance cost becomes a real problem. Before this time you're just adding and adding. After that, you start forgetting stuff you don't use to make room for new stuff. This stuff is either lost again (like muscles) but some abstracted patterns remain making it easy to relearn (just like it's easier to regain strength than to acquire it in the first place). For example, when I started my quant job I was flabbergasted that I remembered practically nothing when it came to taking derivatives (something I was in my humble opinion damn impressive at between ages 16 and 24). It was quickly relearned though---much faster than the time it took to learn the first time around. OTOH, at the age of 35-40 I could do what the youngesters couldn't: write down the solution or something close to it w/o solving for it mechanically. Lots of experience meant that I had a pretty good idea of what it would look like. Indeed, in physics and sword fighting, it's frequently observed that youngsters tend to excel in technical skill and speed whereas oldtimers excel in wisdom (quickly distilling the essence of the problem).

Another thing I've learned in terms of learning many things is that there's an upper limit to how many kinds of expertise it is possible to maintain. Because of that it is absolutely necessary to develop an additional layer of meta-knowledge, somehow. This is again similar to physical development and what crossfit was supposed to solve but the limit seems similar.

Another concern is that the crystallized intelligence gets so entrenched that it's hard to let go of it and rely on fluid intelligence. This kills creativity. I also find myself immediately concluding that "this can't possibly work" more often than I did 20 years ago. I'm mostly right too but not always which causes me to miss some things. I am, for example, not sure I would be able to write the ERE book as it was/is back then now. Creativity requires a certain amount of ignorance. Also, one might not be as eager to explore insofar one has come to believe that there's nothing new to be found.

One can try to deliberate enter new situations to compensate. However, there's also an increasingly lack of new situations as they get replaced by "been there, done that".

Here's a fun thing to contemplate on a societal scale: What if this is the case for scientific knowledge as a whole. We're culturally used to the idea of eternal progress in the same way that a 20 year old still find themselves knowing more and more year after year (based on the previous 20 years of experience)... but what if we're about to turn the corner and there's not much more to know. S-curve! For example, the field of physics hasn't really learned anything mindblowing new for almost 50 years now. It's mostly been refining old stuff. Most of the fluid intelligence was spent on a field that might be useless (string theory).

[Sclass]

@jacob

Hey that was really interesting.

In particular this jumped out at me.

Another thing I've learned in terms of learning many things is that there's an upper limit to how many kinds of expertise it is possible to maintain. Because of that it is absolutely necessary to develop an additional layer of meta-knowledge, somehow. This is again similar to physical development and what crossfit was supposed to solve but the limit seems similar.

Reminds me of physicists working in engineering. Classical wave theory in media for example can get you very far in electronic filter design, optics, acoustics etc.

[daylen]

Here's a fun thing to contemplate on a societal scale: What if this is the case for scientific knowledge as a whole. We're culturally used to the idea of eternal progress in the same way that a 20 year old still find themselves knowing more and more year after year (based on the previous 20 years of experience)... but what if we're about to turn the corner and there's not much more to know. S-curve! For example, the field of physics hasn't really learned anything mindblowing new for almost 50 years now. It's mostly been refining old stuff. Most of the fluid intelligence was spent on a field that might be useless (string theory).

Yeah, it seems the core principles/rules in each discipline/level are solid overall, and there exists a great deal of literature interpreting the core structures/models. The mathematical literature gives a rich library of structures, but there exists a few mathematical barriers that hold science back (especially with computability). Maybe quantum computation can help solve some of these problems and open up room for innovation.

I think that we are near the edge of knowledge that humans are capable of using meaningfully. There are still a lot of details to fill in, but the remaining revolutionary ideas may be few in number. Perhaps "Postmodernism" wasn't such a bad term after all?

How long does it take until the probability that humanity/nature destroys all literature reaches 50% ?! .. in 150 years when a postmodern empire rises by burning everything except Nietzsche and spreading its misinterpretation to a population of eco-technic farmers/scavengers/meat chemists.

[daylen]

I think there is still a huge potential for nano-machines, but the energy required to research may limit nano-tech for now. Such fine-grained manipulation would likely change our society in ways we cannot imagine right now. The line between biological and mechanical would blur.

This may be a good thread to relate to (data, info, know, wise):

Data: List of measurements/observations.
Information: Data with intent.
Knowledge: Information that can accurately deduce the probability of specific measurements/observations (works for all actors at any time).
Wisdom: Ability to use knowledge and awareness of specific context to "steer" the situation in the preferred direction (hence it is based on how the individual relates to their environment).

Over time it makes sense that a human would focus more on high-order data structures. When a human encounters events that disagree with their higher-order frameworks, then the options are to rebuild the framework (consistency + completeness > time), to throw away the data (consistency + time > completeness), or to create a new branch of interpretation in the framework (completeness > consistency).

This is not the complete story, since acting in a wise way also requires awareness of some specifics to tune model parameters. Reality is infinitely differentiable, so acting wise requires a repository of knowledge or meta-knowledge frameworks that can quickly direct attention to the relevant details of a situation as new objectives/obstacles arise.

Knowledge is for cutting; wisdom is how much cutting. By cutting, I mean setting a boundary of some kind (physical, mental, social..). Maybe disagreeableness is how many boundaries and conscientiousness is the average strength of boundaries?

[IlliniDave]

I think what's changed is that as a young man I had an inflated assessment of my intelligence from scoring very high on the various standardized tests and generally being able to out think most people in domains where I focused a lot of effort. Simple youthful arrogance. As I've aged I put less emphasis on more-versus-less and more on recognizing what's simply different. Gradually swallowing a dose of humility over the decades caused me to more fully acknowledge my weaknesses. Mentally there are some things I'm as strong at as I've ever been. But there's things I suck at and always have. Rote memorization is an example. If my aptitude had aligned with my deepest interests I might have wound up in biology, but the rote memorization required even in middle/high school nudged me away from that. There are some artistic pursuits in which I have some ability, others where any foray in their direction was catastrophic. The process of gradually owning shortcomings over time sometimes make me feel like I'm "losing it", but I don't think that's the case yet.

And that doesn't even touch on the topic of wisdom. I'm far, far more able now to synthesize knowledge and experience in modeling scenarios and making decisions that are relatively optimal. It just requires discipline to sidestep the natural tendency humans have to habituate.

Someday the physiological mechanisms behind mental activity could begin to wane, but either that's not started to happen yet, or the physiological mechanisms are not and have never been limiting factors for what I've tried to wrangle from my mind.

[jacob]

@daylen - But there is likely a limit to the achievable order of these data structures:

Going with cybernetic model, we have
1) Being able to respond.
2) Being able to modify response (acting on feedback)
3) Being able to modify the feedback based on response (this is called training)
4) Being able to modify training based on feedback (this is called learning)
5) Being able to modify learning based on feedback (this doesn't really have a name, but is usually referred to as paradigm-shifting, mind blowing, meta-cognition, ...)
6) Being able to modify that based on feedback (this doesn't even have a reference; we can call it meta squared, but that doesn't really tell us anything).
and so on ...

1 describes the intelligence level of a tree; 2 of many animals and most human made/mechanical devices, like e.g. a thermostat; 3 of standard AI, smart animals, and many humans; 4 of most humans; 5 of what is likely the human limit at least as judged by what language constructs we have. Of course we can abstract this idea infinitely high, but we don't have any words or experience to describe it.

One might now wonder what kind of thoughts a tree is unable to have because it's at level 1. Similarly, what is the limit to what humans can think about because we're sitting at level 4-5. Therein lies a likely constraint to human thinking.

Another way to put it is that while it is easy to have ideas (4), it's somewhat rarer for humans to have ideas about ideas(5), whereas having ideas about ideas about ideas(6) is not a thing.

[ThisDinosaur]

Going with cybernetic model, we have
1) Being able to respond.
2) Being able to modify response (acting on feedback)
3) Being able to modify the feedback based on response (this is called training)
...
1 describes the intelligence level of a tree; 2 of many animals and most human made/mechanical devices, like e.g. a thermostat; 3 of standard AI, smart animals, and many humans;.

Nitpicking, but I'd put trees at 2 or 3 here. Even a single cell is at least a 2.
https://en.m.wikipedia.org/wiki/Transcription_factor

[jacob]

Sure, but in that case, I'd put a rock at level 1 because it responds to gravity. It comes down to the definition of "respond".

[ThisDinosaur]

Yeah, well, words can be very ambiguous. But biochemistry, especially at the level of functional genetics, works very much like a computer at the level of logic gates. Cells sense an environment, perform computations, and initiate responses.
Any cell has to be able to "modify responses" based on environmental input in order to qualify as living.