Comment by rdevilla
12 hours ago
I wonder what kinds of information are worth keeping resident in human carbon wetware, and what kinds of information are better off sitting in a silicon cache somewhere in the cloud. On one extreme LLMs do 100% of your thinking, and your brain understands nothing other than how to function as a transport layer from/to the data center and other humans. On the other you have the technophobic tendencies of Anathem's avout that eschew technology in favor of the development of the natural (vs. artificial) mind. It's not clear to me how to carve up the varying cognitive responsibilities between man and machine.
> He warned against mistaking command of words for possession of the solid things those words are meant to disclose. He joined language to substance, sequence to maturation, and study to direct contact with reality — principles that four centuries have not made less urgent.
There are maps that accurately represent a territory, and purely fictitious maps with no relation to any territory whatsoever. This is the spectrum of representation, and LLMs are pushing us towards creating maps that overwhelmingly occupy the latter extremity.
> More writing done in class. More oral defense of arguments. More seminars organized around live questions rather than passive downloads of information.
It's one thing to memorize arguments in favour of a position. It's another to actively defend your positions against those aggressively invested in proving you wrong. John Stuart Mill argued that only the latter activity produces the real understanding that allows an argument, or a tradition, to be renewed and kept alive across generations against constant attempts at refutation. If you are regurgitating a stance instead of actively fighting to defend one, do you really believe in what you are saying?
I think belief that words accurately represent a reality is going to become increasingly important in the years to come. There are now many pantheons to worship at in the 2026 ecosystem of ~digital gods~ AI models, and the question becomes whose version of reasoning you choose to accept as authoritative. Unfortunately, no single model can itself answer this question for you, for obvious reasons.
> I wonder what kinds of information are worth keeping resident in human carbon wetware
I’ve never been an arts person and I’ve been a very, very logical person, so it’s very odd to me to realize that my answer to this is: poetry.
More and more these days I look for ways to both reason with and frame the world and current events. I’ve followed years and years of people putting forth logic and reason as explanations. But my moments of peace are when I find those perfect words written in some distant past, making me feel connected with others by a timeless dimension
Welcome to the aesthetic world! In the western philosophical and certainly scientific discourse there has since centuries been this drive for objectivity and universals. This has led to great discoveries and thinking. But it’s not the only world, the aesthetic is all about the senses and your place as a subject. It usually invites relativism, sometimes nihilism if you can’t find your ground as an individual in a larger universe.
The world of beauty, art, peace, feeling states is worthy of discovery and like you say, it has a timeless quality.
That’s one good welcome! Even I feel welcomed and I have been hanging out in the music section for ages. Other than the music though I can relate to being a logical/rational person.
The majority of poetry is the equivalent of slop created to get into someone's pants. And then there's Pessoa.
Repetition of basic knowledge is actually a big part of a successful education, Even schoolkids in the earliest grades can actually learn surprisingly complex subjects by heart simply by blabbing everything back word-for-word. Problem solving skills can then be built up on these basics.
I am not convinced by that. Kids tend to learn problem solving (and other) skills if given a chance. i do not think encourage huge amounts of rote learning is an optimal, or even, useful say of doing that.
My experience (with myself and my kids) has been the opposite.
We used to have these questions about "What are the advantages and disadvantages of X?"
I used to think I was outsmarting "the system" by only learning a few key facts about X and then twisting them around to get advantages and disadvantages, but little did I know that was the whole point of the course — to see the same thing from different perspectives and realize there are both advantages and disadvantages to X.
https://en.wikipedia.org/wiki/Multiple_drafts_model
Daniel Dennett talked about this model of consciousness. Something similar could be replicated by AI's own self-play style reasoning. It could sharpen its own drafts. As new data points become available, the drafts could be extended, shelved, or reformulated. AI could make these notes on its own reasoning available for others for inspection and course-correction and avoiding local-minima. Common objections need not be raised by layman, AI can incorporate those by itselfs. The true feedback quality can only come from experts in their domain. Implications are what role do normal non-experts have when AI can do most of mid-to-expert level thinking on its own. Hopefully it could help students reach expert level faster.
> It's one thing to memorize arguments in favour of a position. It's another to actively defend your positions against those aggressively invested in proving you wrong. John Stuart Mill argued that only the latter activity produces the real understanding that allows an argument, or a tradition, to be renewed and kept alive across generations against constant attempts at refutation. If you are regurgitating a stance instead of actively fighting to defend one, do you really believe in what you are saying?
A person generally cannot effectively, fluently, convincingly regurgitate an argument without understanding it, and the act of memorizing a variety of different positions primes the brain to handle all of them with greater depth and adroitness. Mill greatly underestimates the power and benefits of memorization.
I think most people would agree that memorization and a standarized 'one-size-fits-all' approach are inferior to teaching methods that are (onstensibly) creative, 'active,' and individualized.
I couldn't disagree more strongly. It's a false dichotomy. All learning -- all -- starts from and depends upon memorization. Is that its only the goal? Obviously not, but memorization gets a bad rap because it's viewed, incorrectly, as contrary to or in competition with more active, creative intellectual enterprises.
I once heard a lecture by a (famous) college professor who talked about the large numbers of students who failed (college) Algebra 1.
His argument was: you cannot memorize algebra, you have to understand. Students who are failing in college do so because they do not understand the fundamentals, and try to memorize enough to succeed - not realizing that the effort needs to go somewhere else.
Rule 1 of memorization is "do not [memorize] if you do not understand". [1] (Note: that source uses the word "learn" instead of "memorize", but to me the word learn means come to understand.)
There is a role for memorization and rote repetition, but it is not the foundation of understanding.
[1]: https://super-memory.com/articles/20rules.htm
As a teacher, I feel this is wrong. A lot of students fail by trying too hard to understand.
They listen in class, then read the text and notes you posted, then watch a Youtube explaination, then ask Chat, then ask you questions ... anything to avoid trying to do a few practice questions where they might make a mistake.
It's like watching people try to learn to play basketball when they are afraid of shooting hoops in case they miss a shot. So they watch videos or read books to really understand how to shoot hoops. And then fail miserably when they are tested.
OK, you could argue that exercises build a type of understanding, and listening to explanations builds a different type of understanding, and the former is more useful, but people don't understand that.
Yeah, memorization is very underrated.
Memorization increases the size of the building blocks you can use.
Mathematics is where I see this most clearly. Why memorize hundreds of theorems? Because then you can just cite them on the fly when doing real mathematics. If you had to re-derive everything, you'd be stuck doing undergrad level math forever.
Chess Grand Masters have large repertoires of memorised openings. They do not play rote games with no understanding.
They run variations, twists and traps, on recalled openings and duel and fool by creating and breaking expectations.
In line with a number of other activities rote core skills and reflexes are foundational but not all, they're essential to practice and to dealing with situations where they don't fit but can be bent to purpose.
> Chess Grand Masters have large repertoires of memorised openings. They do not play rote games with no understanding.
This is a good example because a test of one's understanding is "do you know how to make the opponent pay for varying from the standard opening?"
For a beginner, the answer is no.
> accurately represent a reality is going to become increasingly important
I personally perceive a decoupling all over the board. Not just in language. You hear terms like "wage stagnation" or "degree inflation". Just choose an area. They're all detachments from the true thing they represented.
> I think belief that words accurately represent a reality is going to become increasingly important in the years to come.
Confusion between words and reality has been an important aspect of all human cultures since there were words. It's one of two traditional forms of magic found everywhere. (The other being sympathetic magic.) Think about what it means to say "knock on wood".