Pessimism about Techniques in the Face of AI

General Memory Chat
1

The last few years, I have intensely studied memory techniques, learning methods, and other systems that had apparent potential to improve human mental abilities. This was all in an aim to eventually produce my own system that merged them all together in an effective way, such that it would make learning dramatically faster for humans. This was something of a dream.

But something new is on the rise: AI, showing amazing potential and exponentially improving. OpenAI has proved hundreds of significant open conjectures in mathematics with AI, and even in my personal mathematics research the newest models can solve conjectures from my previous papers in minutes.

It seems human intelligence will be rendered basically insignificant in the face of AI, which will be (and in some ways already is) way faster and more efficient than even the smartest humans. If this is all true, what is the significance of making learning techniques for humans?

Even a successful system will be little more than a novelty in a year or two in the face of an agent swarm of 10000 coordinated Opus 7.5 AIs with dramatically cheaper tokens. What will be the point of human intelligence optimization?

I think I studied the wrong field these past few years: I should have focused on optimizing machine intelligence, not human intelligence. Now I don’t know what to do with myself seeing as though my dream has gone the way of the dodo.

2 Likes
2

AI is a bad facsimile of the human capacity for imagination, creativity, and capacity. Until we break through the next barrier into General Intelligence, it’s just a factory of millions of monkeys doing the work that no one person has the time to do themselves. Yes, including solving math problems that require tens of thousands of human hours of work to solve.

Every time humans invent a system that makes their lives easier we choose to compromise our ability to perform a task our bodies have evolved to do. And, often, this means we have to compensate by supplementing an artificial version of that task into our already busy daily lives.

We invent writing, so we don’t have to remember oral narratives, but then we have to carry scrolls around with us any time we want to remember something. We invent cars, so we don’t have to walk everywhere, but then we need to go to the gym and get on treadmills. We invent the internet so we don’t have to leave the house to buy things, but then we have to scroll facebook for hours in order to feel a sense of community.

Some developments are neutral, some are positive, and some have an overall negative effect on the world. We invented atomic bombs so we’d never be threatened, but then everyone got them and we lived in constant fear.

LLM-chat and generative AI promises to make it so we don’t have to think for ourselves, and can outsource the one thing that’s always been our own, our minds, to companies that bill us a subscription fee for it. It’s the yawning maw of oblivion, offering to let us drift into the void of obsolescence for a moderate monthly fee. There is no reason, in fact, there is no justification to walk willingly into those teeth.

Unfortunately, many will, which means that what you do, optimizing your mind, is now a premium ability. Unappreciated by many, but vital to others. It cannot be replaced, not by a billion monkeys on a billion typewriters for a trillion tokens.

16 Likes
3

I don’t know. Ai can’t write a large book even. If you don’t have studies you have to do low paid jobs. .

1 Like
4

I think there are still many flaws in AIs, and for certain mathematical riddles—even though I am not from the field—AI can create solutions that we lack the equipment to verify whether they are correct. But even if artificial intelligence becomes superior to human intelligence, if we don’t exercise our brains and bodies, they will simply atrophy. I have a strong impression that technological conveniences will be an additional factor in the rise of Alzheimer’s cases. Using memorization techniques brings me happiness, greater attention to details every day, and looking with curiosity at each piece of knowledge I place in a Memory Palace, for example. Because of this, I become more present in every moment and absorb more from meeting people—whom, by the way, I don’t forget the names of five minutes after being introduced. It may be a moment of career transition for you, and that is a great learning experience. But AI cannot remember for me, a computer cannot eat for me, and capitalism itself still needs real people consuming. AI may well be superior to humans in many aspects and replace humans in many ways; that is Darwin’s law. We can also learn to accept this fact, just as many animals have certain abilities that are far superior to ours as humans.

Anyway, just a reflection open to critique upon reading your message.

4 Likes
5

Spoken so eloquently and truthfully

4 Likes
6

There is when you are facing people using those teeth to 100x their abilities. You will be crushed like a fly going against them. It will be like running at an F1 race.

There is something to be said about continuing to exercise natural abilities, though. But why would we need memory techniques for this if our only goal is not to atrophy?

It can easily, it already could in 2024 when amazon was being flooded by them. For example, many of the memory guides you find on amazon are AI generated. Recently, a professor put out a 20-volume textbook on QFT using AI.

As for studies being needed for high-paying jobs, it is very unclear what will still be of value after four more years of exponential AI improvement. It’s no longer a clear inference to make into the future.

This is not true, AI can easily create LEAN verifications of proofs.

And no, being forced out of the job you’ve been training for for six years is not a great experience, especially for someone with a research background in our hyper-credentialized modern world during a time of rapid change.

Ditto for “There is something to be said about continuing to exercise natural abilities, though. But why would we need memory techniques for this if our only goal is not to atrophy?”

…

Funnily enough, just a day after I posted this, OpenAI released 700+ manuscripts containing 300+ proofs to major open problems, each oneshotted and only being allotted 3 hours of thinking time. Such proofs include the quasi-Riemann hypothesis, Hilbert’s tenth over the rationals, and faster matrix multiplication.

I think the logical response to these developments is to abandon learning techniques and to double down on machine learning knowledge. What do you guys think?

1 Like
7

Hard disagree. You’re right that the best AI tools can do these things, and those capabilities are rapidly growing. But I really don’t care if someone can “100x their abilities.” I can “100x” my ability to travel quickly by getting on a jet. I can move 100x my heaviest barbell using a forklift. That doesn’t mean I shouldn’t jog or do weight training.

An AI can write a million poems in the time it takes me to write one, but those poems are meaningless to me. An AI can write code thousands of times faster than me, but unless I have a QA pipeline that can handle that output, all I’ve done is max out my bottleneck. It’s like installing nitrous tanks on a golf cart.

In other words, I’m not going to try to compete with people who compromise their intellect with AI, because productivity has never been my goal. It has its place in medical tests and cutting edge math and so on, but it would add nothing to my life or my work in IT. I’ve actually wasted many hours trying to parse incorrect troubleshooting information given to me by coworkers who used AI to research and write it- and I can’t think of once when it actually helped.

Lastly, what is “machine learning knowledge?” Do you mean researching how AI can be used in agent swarms, how it works on a GPU, what it’s made of? Sure, that’s someone’s job. Hats off to them but it wouldn’t help my life. But if you mean learning how to talk to a chat bot, I’m not going to waste my time. Not only is it incredibly easy by design, not a skill, but it’s going to continue changing rapidly and get easier and easier. “Prompt sculpting” or whatever people call it is not only not valuable, it’s probably obsolete by the time you read this.

Unless I needed to use AI as a way to expand the boundaries of my field of work like in medical testing, I don’t see what it could possibly do in my life other than supplement my capabilities into a dark and crippling dependency,

9 Likes
8

my two cents: i think AI gives us the opportunity to learn in fascinating and exciting new ways. i suppose if you want to use this new technology instead to avoid learning that’s also easier than ever, but i think that’s silly.

1 Like
9

We have to use it because it’s possible all jobs use it in the future. But is possible that agents can do our own job too. And almost free.

1 Like
10

The inevitability of AI is a narrative AI companies have spent fortunes on. The truth is it is not inevitable and you do not need to use it.

5 Likes
11

I’m not sure. Now ai is used by people to make CV to look for work . And the CV is read by other ai before a human can see it. Therefore unless you just got a job now you need to know about this technology. Yes is true 20 corporations are wasting billions on it and can stop any day like with .Com enterpises happened once. But today it don’t seem likely.

1 Like
12

You have to admit that “jogging” and “weight training” are way less significant than they were before the creation of the tools you mentioned. This is exactly my point: I wanted to do something significant with these techniques, more than bolster a surrogate activity (which is essentially what jogging and weightlifting have been reduced to).

The newest models are not even comparable to those out a month ago ESPECIALLY in research. FAANG companies are almost exclusively relying on AI code now, and many software engineers haven’t touched code in a while.

As for machine learning knowledge, I mean the former. Local models are becoming better and better, and AI capabilities might drive GPU prices even higher in the future. The latter (“prompting skills”) seems like a joke, unless you mean learning about skills and things like that which can have value.

Also, you should care about whether someone can 100x their abilities. Your job existing hinges on facts like this.

You can try to spin it like it’s just something pushed by AI companies, but you’re ignoring the fact that unreleased models can solve quasi-riemann hypothesis in three hours. The new Opus 5.5 models can already crush conjectures us masters students were working on. Do you think it will just stop at being better at math than mathematicians? That is an awfully convenient viewpoint to have, don’t you think?

1 Like
13

I don’t understand where this pessimism is coming from. Unless you expect humanity to not exist in the coming future, you’ll still need to rely on your memory to do some things. Since we’ll always need to rely on our memory for some things, memory techniques will always have some relevance.

7 Likes
14

What about when high-bandwidth brain-computer interfaces are here? Our thinking would be offloaded to machines.

15

In that hypothetical future, mnemonics probably would become obsolete. But I think we’re quite a ways off until then.

16

I agree they will have some relevance, but you must admit their significance (as well as learning techniques) will take a big hit when thinking and research is largely directed to llms. I made a similar point above with running before vs after more efficient methods of cheap transportation. I agree they will still be a hobby and hold a bit of significance, but on the whole they will go the way of the dodo.

17

No, I don’t think their relevance will decrease. So long as humans need to memorize or learn anything at all, memory techniques will be of some importance.

Edit: If by importance you mean popularity, then I would agree. But just because something can be important/significant to one’s life doesn’t mean that it will popular with the broader public.

2 Likes
18

Maybe. I’m not denying the capabilities of AI, or the exponential speed at which it is evolving. I just believe its pragmatic usefulness is so far pretty universally overstated, and in terms of the value consumer-level chatbots bring, by orders of magnitude.

I feel that there’s a degree to which we’re describing the same phenomenon accurately from our perspectives but we’re disagreeing because we’re making predictions about something unprecedented and about which we have no prior experience. So it’s easy to come to different conclusions because there’s very little evidence for us to examine.

The perspective I’m coming from is that in my personal and professional sphere I have not seen AI produce anything useful. In fact, it has made my job in IT much harder, with exploits and viruses being written a million times a minute, and microsoft pushing out buggy patches in response, also written by AI and buggy as hell. Evidence points to the conclusion that chatbots are dangerous for even mentally well people, causing psychosis and delusion, as well as making them less intellectually capable when not augmented by their favorite chatbot.

In short, while I’m glad it exists to further medical fields and other vital research areas, I’m not worried about it taking over my job personally nor am I interested in using it. I’ll learn about it (and have been for years) but the skills I value in my life cannot be replicated by anything without a human body.

4 Likes
19

Eventually grandmothers will be retrofitted with tires and then they will be bicycles. But we’re thinking far too ahead for our lifetimes.

3 Likes
20

You hit the nail on the head, though. Companies are using AI hiring agencies to read AI-written CVs and everyone is wasting time and money except the AI companies who are making revenue (not profit) hand-over-fist. As long as you can get a job some other way, I think you should. And if you are hiring, don’t waste your time and your applicants’ time with unhuman hiring agencies who aren’t going to be able to do the work a good hiring team can.

4 Likes