7 ms·

Zen and the Art of Machine Learning Research

289 points by jxmorris12 4mo ago

lostdog 4mo ago
I have some coworkers that are similar in everything--education, work ethic, and intelligence--but some of the tick out ML ideas that work like clockwork, while others get hits rarely if ever. I cannot tell what makes it work for some and not others. Their ideas both sound equally good. Sometimes a coworker will be an ML star for a year or two, but then suddenly run out of steam. It's brutal to watch. I used to think most smart people had similar distributions of good ideas, and it was just that the hardest working tried out all 50 of their ideas to pick out the 2 good ones. But I've seen smart and hardworking people have a hit rate of 0.
- fyredge 4mo ago
  That's the nature of research. You try every idea that may be a good avenue and only a handful work out, if at all. That's why quantifying research credibility via publication and citation counts inherently lead to toxic work cultures. The best ideas must be given time to be discovered, not forced out and contorted to fit the requirements of a journal.
  - bobmarleybiceps 4mo ago
    this is part of why I think most researchers get less productive over time... Someone gets some big result during grad school or early career, get some big job from it, and then struggle to get new results of similar quality :shrug: With ML in particular, there's also the sheer volume of people basically all looking at (essentially) the same problems... so it's kind of like monkeys with type writers spamming ideas until some work.
- jack_pp 4mo ago
  In spirituality it is believed that ideas and inspirations aren't our own. That our mind is like an LLM that gets prompted by higher beings. In research everyone has high param count minds, trained for many years by studying. But just like LLMs by themselves are useless at creating new original work, no matter the compute you have available, so the mind can not create anything new without "inspiration"
- 59nadir 4mo ago
  Wow, this makes ML sound even more like voodoo than I thought. Can you give examples of what the nature of these ideas is?
- sdsdfsdff344sd 4mo ago
  It's not just ML research; that's just human nature. We like to see hard-working, God-fearing people minting raw knowledge from Mount Olympus itself, whereby each shard of crystalline insight is carved meticulously by the Apprentice over the course of a productive and morally pure career. The reality is it's some skill plus the occasional drive-by of an unknown force of nature, hitting you on the head with a shattered fragment of insight whose provenance you'll remain completely ignorant of. I'd say we just revert back to invoking the muses. It was a fine explanation.
- cold_harbor 4mo ago
  [dead]
jdw64 4mo ago
I feel that the Zen used in the West and the Zen in East Asia are quite different. I think the Western Zen is probably the one from the 1970s book Zen and the Art of Motorcycle Maintenance. It usually carries a sense of equanimity and beginner's mind. But in East Asia, Zen actually emphasizes aimlessness or non‑purposefulness. The point where I really feel the difference is that Western Zen seems to be about how to train the self to become stronger, whereas actual Seon (Zen) in East Asia is about going with nature, letting go of the self, and allowing things to flow. In the actual practice of Seon, it's about doubting the self, letting go of attachments, and realizing that achievement, comparison, and the desire for control are all just fleeting. There's a famous phrase: 'Banghasak (放下著)' — let it all go. If anything, I think ancient Roman Stoicism feels more like Zen than Western Zen does So that's fascinating. When I saw this article, I was expecting it to be about whether we should give up the desire for success, but instead it took a completely different direction, which was surprising
- peepee1982 4mo ago
  Similarly, the Western idea of Stoicism seems to focus mostly on controlling or even suppressing your emotions (at least on surface level), while the Stoicism you rightly call "Roman" (thanks for that, btw) is much more holistic and more of an ethical framework.
  - jdw64 4mo ago
    Thank you for letting me know correctly.
    - oliculipolicula 4mo ago
      Greek Stoicism (no quotes) is more positive. ὁ ἀγαθὸς ἀνὴρ [ἀδιάπτωτος καὶ] ἀδιάσειστος The good man (agathos aner, man who acts for the benefit of others) is [infallible and] unshakeable. --Epictetus, whose name sounds "Roman", just like "Stoa" doesn't This can be (plausibly) used to describe Sam Altman, for example
      - 4mo ago
sdfsefsdf 4mo ago
Perhaps I've been deep in my own issues for too long, but it seems to me that the author is trying to say "don't trust the current evaluation suites too much"; scores only reflect a small part of the problem. What's interesting is discovering a new, stable evaluation metric, doing something new based on it, and having that new thing yield some unexpected intelligent results
- jxmorris12 4mo ago
  This is certainly part of it! My point was that focusing on problems proposed by others is one very specific and pretty short-term mode of thinking. Good researchers improve benchmark scores. Great researchers think about what problem they're solving.
nathaah3 4mo ago
This is gold!!!!
rented_mule 4mo ago
Around 2015, I found myself managing back end and machine learning engineers (not researchers) at the same time. Many of the back end engineers wanted to do more ML. Some of them did well when given a chance, but others wanted to revert to back end within a few months. At the same time, one of the ML leaders wanted to step away from ML and only do back end work to support ML. As I studied these dynamics, something occurred to me... Different people need to see signs of success at different frequencies. Because of the nature of our product, measuring the performance of a new/updated model required the model to be live for at least a full calendar month. So, between initial work and final analysis, it was often a 2 month wait or more. For many back end tasks, you can build a quick prototype, run it to see if it works, and be on your way - the signals come all day long. The varying frequency needs of different people went a long way to determining which of them liked working on ML. This is sort of a manager's version of feature engineering. ;-) The people on that team taught me a lot!
- dalvasorsali 4mo ago
  I saw the same thing and always wondered how you can manage it effectively. I had a team of data engineers that wanted to do more data science, and 2 data scientists that both wanted to be data engineers(one of them argued that everyone wants to be DS and so it was too crowded, saying that they could make more money as a DE). I also remember a specific instance where, one day, my friend ranted about how he needs to step away from pure front end and that it's a dead end career (he was quite good at it too!) and then the next day at lunch a colleague started complaining about how front end developers get all the credit and he's considering moving.
  - rented_mule 4mo ago
    Sometimes it's a "grass is greener on the other side of the fence" thing or a FOMO thing. But sometimes different roles fit different people better. I started my career doing mostly full stack work. I couldn't get away from the front end part quickly enough. I had intuition for simplifying UI flows, but none at all for aesthetics. As requests for aesthetic changes came in from our excellent designers, they felt completely arbitrary to me, even though they probably weren't. Most of my career was as a data engineer, data engineering manager, or leading an ML-heavy org. That space fit me so much better. I loved having a few self-starting front-end devs in my orgs - they could take various tools we were creating for ourselves and make them quite a bit more useful. But it was also always a stepping stone as they typically wanted to work on the public facing part of the product.
stared 4mo ago
It revolves around the sentiment of "go deeper" - but I think it is a double-edged sword. Sure, entropy, tensors and gradients are important - and yes, they are pretty much requirements. But from what I see, it is the opposite - a lot (if not virtually all) progress in the last decade of deep learning was not because of a fundamental idea, but incremental, experimentally-verified practice. Even though I think there is good intuition for why ReLU is better than sigmoid (tl;dr: last layer is log(sigmoid) ~ ReLU, putting anything different inside kills the gradient), the original paper by Hinton himself was more or less "because it trains 3x faster". Re-thinking fundamentals might help, but most "let's change the fundamentals" is rarely how it works. Even the most seminal papers, i.e. AlexNet and "Attention Is All You Need", are refinements of existing ideas, and show how they help. Machine learning is an experimental science. Many mathematically cool ideas do not work. Many engineering ones do. > I've tweeted before that one of the most important traits in a researcher is healthy paranoia. Be paranoid! I have seen so many PhDs burned out to cinders; I don't think it is any more a good piece of advice than "depression is good for philosophers". Sure, be a relentless explorer. > In short, holding on to ideas for too long can actually be counterproductive. Stay open-minded and refuse to let ego cloud your judgement. Which I think is true.
WithinReason 4mo ago
> If you want to solve a problem, the tried-and-true path to success is to attempt a solution, try it, reach a bottleneck, try to solve it, and only reach for literature when you’ve run out of ideas yourself. I've found this to be the right balance between using your creativity and getting stuck too long
Scene_Cast2 4mo ago
I think this also stems from ML being more like biology or alchemy and less like math or programming (where you can get down to the first principles, abstractions are rock solid, and non-determinism is limited in scope).
mrmarket 4mo ago
excellent essay. what a great read. like the author said, so much of 'success' or 'progress' (in research but of course also across disciplines) depends upon temperament. just straight up having a good attitude about things. the skills that make a good researcher could not be more transferable: patience, innate curiosity, and a resilience against failure. that said, these skills are increasingly rare/at a premium given our culture of minimizing discomfort tolerance via hyperconvenience. people have a harder and harder time waiting or failing.
almarcher 4mo ago
Stepping away from the work to find inspiration, to allow the subconscious time to process everything, to present your conscious mind ideas is necessary. I try to pick a wild or almost outlandish idea from time to time, because if I only try what I think will work, then I'm not doing my job.
HarHarVeryFunny 4mo ago
> on days we find insight, we sit. > on days we do not find insight, we sit. This reminds me of Ed Witten (greatest living physicist?) in an interview by Brian Green. Green asked Witten what his day-to-day was like at the Institute for Advanced Study ... Wittens' reply: "I sit at my desk".
jessinra98 4mo ago
Would either of you have a recommendation on where to start learning about either?
aputsiak 4mo ago
You can in fact take courses taught by the greatest in the field. The one does not exclude the other.
photochemsyn 4mo ago
[flagged]
misiti3780 4mo ago
why is SVD so important? i know it's important in general ML but seems minor for LLMs (LoRA?)
ms_by_pd 4mo ago
Good learning
hottrends 4mo ago
[flagged]
strix_varius 4mo ago
> I’ll level with you – this is hard! I'm getting tired of the monoculture where both articles I read and agents I code with speak with the same (annoying) voice.
- jxmorris12 4mo ago
  Haha. Unfortunately is my regular voice, since long before I started using Codex. You can check through some of my old writing. It definitely could've gotten worse though. Not sure if I'm training on Codex, or Codex is training on me...
jjssmith 4mo ago
> One really impressive thing about OpenAI is that most of the people running the company (on the technical side, at least) are under 35. Many of the important decisionmakers behind chatGPT are under 30. During Gold Rush most 49ers were under 25, so there's still room for improvement! [Continuing the analogy, you may find that many AI heroes are just those who happened to be closer to pools of TPUs and GPUs in the early days...]
- kilpikaarna 4mo ago
  > [Continuing the analogy, you may find that many AI heroes are just those who happened to be closer to pools of TPUs and GPUs in the early days...] And this in turn was much more likely if you were inexplicably obsessed with ML/AI at a point when it was widely considered unfashionable.
exabrial 4mo ago
On a similar note, I think it's pretty hilarious (short sighted) of Anthropic to have open hiring positions, but ban the use of it's product for... frontier model research. Where do you think these people are going to come from? Rash decision, and will likely draw an anti-competitive lawsuit at some point.
smath 4mo ago
Tangential tidbit about etymology of the word Zen: Zen is a Japanese word that comes from the Chinese “Chan”, which in turn comes from the Sanskrit word “Dhyana”, which roughly translates to focus/meditatiin. That trajectory Sanskrit —> Chinese —> Japanese reflects the geographical trajectory of the spread of Buddhism out of India. Same word in Vietnamese and Korean is “Thien” and “Seon” respectively.

289 points by jxmorris12 4mo ago

I have some coworkers that are similar in everything--education, work ethic, and intelligence--but some of the tick out ML ideas that work like clockwork, while others get hits rarely if ever. I cannot tell what makes it work for some and not others. Their ideas both sound equally good. Sometimes a coworker will be an ML star for a year or two, but then suddenly run out of steam. It's brutal to watch. I used to think most smart people had similar distributions of good ideas, and it was just that the hardest working tried out all 50 of their ideas to pick out the 2 good ones. But I've seen smart and hardworking people have a hit rate of 0.

fyredge 4mo ago
That's the nature of research. You try every idea that may be a good avenue and only a handful work out, if at all. That's why quantifying research credibility via publication and citation counts inherently lead to toxic work cultures. The best ideas must be given time to be discovered, not forced out and contorted to fit the requirements of a journal.
- bobmarleybiceps 4mo ago
  this is part of why I think most researchers get less productive over time... Someone gets some big result during grad school or early career, get some big job from it, and then struggle to get new results of similar quality :shrug: With ML in particular, there's also the sheer volume of people basically all looking at (essentially) the same problems... so it's kind of like monkeys with type writers spamming ideas until some work.
jack_pp 4mo ago
In spirituality it is believed that ideas and inspirations aren't our own. That our mind is like an LLM that gets prompted by higher beings. In research everyone has high param count minds, trained for many years by studying. But just like LLMs by themselves are useless at creating new original work, no matter the compute you have available, so the mind can not create anything new without "inspiration"
59nadir 4mo ago
Wow, this makes ML sound even more like voodoo than I thought. Can you give examples of what the nature of these ideas is?
sdsdfsdff344sd 4mo ago
It's not just ML research; that's just human nature. We like to see hard-working, God-fearing people minting raw knowledge from Mount Olympus itself, whereby each shard of crystalline insight is carved meticulously by the Apprentice over the course of a productive and morally pure career. The reality is it's some skill plus the occasional drive-by of an unknown force of nature, hitting you on the head with a shattered fragment of insight whose provenance you'll remain completely ignorant of. I'd say we just revert back to invoking the muses. It was a fine explanation.
cold_harbor 4mo ago
[dead]

jdw64 4mo ago

I feel that the Zen used in the West and the Zen in East Asia are quite different. I think the Western Zen is probably the one from the 1970s book Zen and the Art of Motorcycle Maintenance. It usually carries a sense of equanimity and beginner's mind. But in East Asia, Zen actually emphasizes aimlessness or non‑purposefulness. The point where I really feel the difference is that Western Zen seems to be about how to train the self to become stronger, whereas actual Seon (Zen) in East Asia is about going with nature, letting go of the self, and allowing things to flow. In the actual practice of Seon, it's about doubting the self, letting go of attachments, and realizing that achievement, comparison, and the desire for control are all just fleeting. There's a famous phrase: 'Banghasak (放下著)' — let it all go. If anything, I think ancient Roman Stoicism feels more like Zen than Western Zen does So that's fascinating. When I saw this article, I was expecting it to be about whether we should give up the desire for success, but instead it took a completely different direction, which was surprising

peepee1982 4mo ago
Similarly, the Western idea of Stoicism seems to focus mostly on controlling or even suppressing your emotions (at least on surface level), while the Stoicism you rightly call "Roman" (thanks for that, btw) is much more holistic and more of an ethical framework.
- jdw64 4mo ago
  Thank you for letting me know correctly.
  - oliculipolicula 4mo ago
    Greek Stoicism (no quotes) is more positive. ὁ ἀγαθὸς ἀνὴρ [ἀδιάπτωτος καὶ] ἀδιάσειστος The good man (agathos aner, man who acts for the benefit of others) is [infallible and] unshakeable. --Epictetus, whose name sounds "Roman", just like "Stoa" doesn't This can be (plausibly) used to describe Sam Altman, for example
    - 4mo ago

sdfsefsdf 4mo ago

Perhaps I've been deep in my own issues for too long, but it seems to me that the author is trying to say "don't trust the current evaluation suites too much"; scores only reflect a small part of the problem. What's interesting is discovering a new, stable evaluation metric, doing something new based on it, and having that new thing yield some unexpected intelligent results

jxmorris12 4mo ago
This is certainly part of it! My point was that focusing on problems proposed by others is one very specific and pretty short-term mode of thinking. Good researchers improve benchmark scores. Great researchers think about what problem they're solving.

nathaah3 4mo ago

This is gold!!!!

rented_mule 4mo ago

Around 2015, I found myself managing back end and machine learning engineers (not researchers) at the same time. Many of the back end engineers wanted to do more ML. Some of them did well when given a chance, but others wanted to revert to back end within a few months. At the same time, one of the ML leaders wanted to step away from ML and only do back end work to support ML. As I studied these dynamics, something occurred to me... Different people need to see signs of success at different frequencies. Because of the nature of our product, measuring the performance of a new/updated model required the model to be live for at least a full calendar month. So, between initial work and final analysis, it was often a 2 month wait or more. For many back end tasks, you can build a quick prototype, run it to see if it works, and be on your way - the signals come all day long. The varying frequency needs of different people went a long way to determining which of them liked working on ML. This is sort of a manager's version of feature engineering. ;-) The people on that team taught me a lot!

dalvasorsali 4mo ago
I saw the same thing and always wondered how you can manage it effectively. I had a team of data engineers that wanted to do more data science, and 2 data scientists that both wanted to be data engineers(one of them argued that everyone wants to be DS and so it was too crowded, saying that they could make more money as a DE). I also remember a specific instance where, one day, my friend ranted about how he needs to step away from pure front end and that it's a dead end career (he was quite good at it too!) and then the next day at lunch a colleague started complaining about how front end developers get all the credit and he's considering moving.
- rented_mule 4mo ago
  Sometimes it's a "grass is greener on the other side of the fence" thing or a FOMO thing. But sometimes different roles fit different people better. I started my career doing mostly full stack work. I couldn't get away from the front end part quickly enough. I had intuition for simplifying UI flows, but none at all for aesthetics. As requests for aesthetic changes came in from our excellent designers, they felt completely arbitrary to me, even though they probably weren't. Most of my career was as a data engineer, data engineering manager, or leading an ML-heavy org. That space fit me so much better. I loved having a few self-starting front-end devs in my orgs - they could take various tools we were creating for ourselves and make them quite a bit more useful. But it was also always a stepping stone as they typically wanted to work on the public facing part of the product.

stared 4mo ago

It revolves around the sentiment of "go deeper" - but I think it is a double-edged sword. Sure, entropy, tensors and gradients are important - and yes, they are pretty much requirements. But from what I see, it is the opposite - a lot (if not virtually all) progress in the last decade of deep learning was not because of a fundamental idea, but incremental, experimentally-verified practice. Even though I think there is good intuition for why ReLU is better than sigmoid (tl;dr: last layer is log(sigmoid) ~ ReLU, putting anything different inside kills the gradient), the original paper by Hinton himself was more or less "because it trains 3x faster". Re-thinking fundamentals might help, but most "let's change the fundamentals" is rarely how it works. Even the most seminal papers, i.e. AlexNet and "Attention Is All You Need", are refinements of existing ideas, and show how they help. Machine learning is an experimental science. Many mathematically cool ideas do not work. Many engineering ones do. > I've tweeted before that one of the most important traits in a researcher is healthy paranoia. Be paranoid! I have seen so many PhDs burned out to cinders; I don't think it is any more a good piece of advice than "depression is good for philosophers". Sure, be a relentless explorer. > In short, holding on to ideas for too long can actually be counterproductive. Stay open-minded and refuse to let ego cloud your judgement. Which I think is true.

WithinReason 4mo ago

> If you want to solve a problem, the tried-and-true path to success is to attempt a solution, try it, reach a bottleneck, try to solve it, and only reach for literature when you’ve run out of ideas yourself. I've found this to be the right balance between using your creativity and getting stuck too long

Scene_Cast2 4mo ago

I think this also stems from ML being more like biology or alchemy and less like math or programming (where you can get down to the first principles, abstractions are rock solid, and non-determinism is limited in scope).

mrmarket 4mo ago

excellent essay. what a great read. like the author said, so much of 'success' or 'progress' (in research but of course also across disciplines) depends upon temperament. just straight up having a good attitude about things. the skills that make a good researcher could not be more transferable: patience, innate curiosity, and a resilience against failure. that said, these skills are increasingly rare/at a premium given our culture of minimizing discomfort tolerance via hyperconvenience. people have a harder and harder time waiting or failing.

almarcher 4mo ago

Stepping away from the work to find inspiration, to allow the subconscious time to process everything, to present your conscious mind ideas is necessary. I try to pick a wild or almost outlandish idea from time to time, because if I only try what I think will work, then I'm not doing my job.

HarHarVeryFunny 4mo ago

> on days we find insight, we sit. > on days we do not find insight, we sit. This reminds me of Ed Witten (greatest living physicist?) in an interview by Brian Green. Green asked Witten what his day-to-day was like at the Institute for Advanced Study ... Wittens' reply: "I sit at my desk".

jessinra98 4mo ago

Would either of you have a recommendation on where to start learning about either?

aputsiak 4mo ago

You can in fact take courses taught by the greatest in the field. The one does not exclude the other.

photochemsyn 4mo ago

[flagged]

misiti3780 4mo ago

why is SVD so important? i know it's important in general ML but seems minor for LLMs (LoRA?)

ms_by_pd 4mo ago

Good learning

hottrends 4mo ago

[flagged]

strix_varius 4mo ago

> I’ll level with you – this is hard! I'm getting tired of the monoculture where both articles I read and agents I code with speak with the same (annoying) voice.

jxmorris12 4mo ago
Haha. Unfortunately is my regular voice, since long before I started using Codex. You can check through some of my old writing. It definitely could've gotten worse though. Not sure if I'm training on Codex, or Codex is training on me...

jjssmith 4mo ago

> One really impressive thing about OpenAI is that most of the people running the company (on the technical side, at least) are under 35. Many of the important decisionmakers behind chatGPT are under 30. During Gold Rush most 49ers were under 25, so there's still room for improvement! [Continuing the analogy, you may find that many AI heroes are just those who happened to be closer to pools of TPUs and GPUs in the early days...]

kilpikaarna 4mo ago
> [Continuing the analogy, you may find that many AI heroes are just those who happened to be closer to pools of TPUs and GPUs in the early days...] And this in turn was much more likely if you were inexplicably obsessed with ML/AI at a point when it was widely considered unfashionable.

exabrial 4mo ago

On a similar note, I think it's pretty hilarious (short sighted) of Anthropic to have open hiring positions, but ban the use of it's product for... frontier model research. Where do you think these people are going to come from? Rash decision, and will likely draw an anti-competitive lawsuit at some point.

smath 4mo ago

Tangential tidbit about etymology of the word Zen: Zen is a Japanese word that comes from the Chinese “Chan”, which in turn comes from the Sanskrit word “Dhyana”, which roughly translates to focus/meditatiin. That trajectory Sanskrit —> Chinese —> Japanese reflects the geographical trajectory of the spread of Buddhism out of India. Same word in Vietnamese and Korean is “Thien” and “Seon” respectively.