6 ms·

I'm probably in the same boat as those people, but in AI. Some of the problems I find interesting are just out of my depth. I don't know exactly what's required

I'm probably in the same boat as those people, but in AI. Some of the problems I find interesting are just out of my depth. I don't know exactly what's required and from where to catch up to attempt a particular question in mind. A PhD student's story on this. https://www.reddit.com/r/MachineLearning/comments/73n9pm/d_confession_as_an_ai_researcher_seeking_advice/ https://www.reddit.com/r/MachineLearning/comments/73n9pm/d_c...

cinntaile 5y ago
The references should allow you to follow some sort of red thread. They will probably aid in closing the knowledge gap as well. It's probably like in this article (it was a good article btw) though, you need more maths. Either you use university curricula to find out what maths you need as a starting point or you focus on the maths that pops up in the papers and work your way from there.
- tomcat27 5y ago
  I can't speak for others, but I find scientific literature is not written general audience to understand, for a variety of reasons such as (a) publisher limitation on number of pages (b) to communicate core message right without distraction (c) to not bore a certain people etc. Not to mention, depending on the sub-fields, at least in AI, there is quite a bit of noise in papers as they are explaining _their_ viewpoint which might differ from broad consensus or disproven in future. I find papers are typically written keeping in mind a specific set of people, who are typically in that particular sub-field for some time. In sub-fields like quantum computation, that specific set of people authors have in mind while writing paper are <100, and those are not newbies. Newbies can quickly get discouraged reading a paper out of the blue. Picking the right papers at the beginning requires exposure to the field. It's not simple. I really hope what you said is practical. I tried. Hope it's my fault because the alternative option doesn't exist for me.
bmitc 5y ago
I think part of the problem is that science is not honest with itself. There is some sense that getting a Ph.D. is an institutionalization process, one that does not really reward non-conservative, long-term, or high probability of failure work. And god forbid if someone wants to revisit "known" ideas and results to help clarify them, apply them, or place them in better context. There is very much an "in" crowd in science, which leaves a lot of people, people who could positively contribute, feeling "out". We have commoditized academic research, at least in the science and engineering fields, which makes things very streamlined and on rails. I've spent some time in both academic and industrial circles of mathematicians, engineers, computer scientists, and physicists. It is a bit astonishing just how people in each one of these groups all sound like each other and distinct from the other groups. There are particular mannerisms and even idiomatic phrases that each group uses. Some simple ones are that physicists like to say "it goes like" and use the word "codes" to refer to code, as in software code. Computer scientists like to say "it better be". There are many others that I've lost track of and should have written down over these years. Yes, in many ways, the little idiomatic phrases make sense, but it at least lends credence to the question that if they all use the same phrases, something very simple and harmless, how else are they "conditioned"? The book Disciplined Minds: A Critical Look at Salaried Professionals and the Soul-battering System That Shapes Their Lives by Jeff Schmidt covers some of this. It all makes me wonder if someone like John Archibald Wheeler, a personal favorite, could be successful in today's academic climate.
- nautilius 5y ago
  The simple answer is: funding. There is no (public) money to revisit "known" ideas and results to help clarify them, apply them, or place them in better context. And I would not be the one to explain to the public to spend their money on known ideas rather than discover something new and shiny. Concerning a PhD, I think it's pretty easy to see that that's not the stage in your career where you would want to embark on long-term, high risk research. Also, I am not surprised that people doing a similar job have a common vocabulary and vernacular. You could say culture. I'd be surprised if carpenters, mechanics, et al. don't have the same thing, with other words and topics, of course. No conspiracies needed. Wheeler had significant contributions, had the right pedigree, the right schools. I don't see how he would be at a disadvantage today.
  - bmitc 5y ago
    It's not about conspiracy theories. It's about reality. You even explicitly called it out: culture. That's just the way humans work, for better or worse. One of my points, and I think a point of Disciplined Minds, is that scientists are sometimes not honest or self-aware of the culture that they're a part of. Science is not a purely rational endeavor. There's some fashion and belief to it. > Wheeler had significant contributions, had the right pedigree, the right schools. I don't see how he would be at a disadvantage today. I didn't mean it as a critique upon Wheeler's potential. He was incredible. But he switched fields and research directions several times during his long and successful career. I'm not sure academics today have that fluidity in their careers (not due to internal forces but rather the external ones).
    - nautilius 5y ago
      I don't see why scientists would be less self-aware of their culture than other groups. In particular, that culture has got nothing to do with the things you decry (why is no one looking at old ideas, where, again, the simple problem is money). As soon as Wheeler would have tenure today, he could switch all day long. I am not talking about Wheeler's potential either, but whether he ticks the boxes to make him likely to succeed today. And he does. You are looking for a career that wouldn't work today? Freeman Dyson. Getting tenure at Princeton aged 29 without a doctorate, not so likely today.
car_analogy 5y ago
> I'm probably in the same boat as those people, but in AI. Some of the problems I find interesting are just out of my depth. You are not. There's an ocean of difference between someone that recognizes their ignorance, and someone that doesn't.
- c22 5y ago
  Surely if someone is paying a consultant to weigh in on their ideas they must have some inkling that they're out of their element. From the article: > One or two seemed miffed that I didn’t immediately exclaim: ‘Genius!’, but most of my callers realised that they can’t contribute to a field without meeting today’s quality standard. Then again, I hear only from those willing to invest in advancing their education to begin with.
abeppu 5y ago
But ... I think with AI, if you have an insight that isn't motivated by math per se, but is motivated by some other problem intuition, you can still attempt to implement it, and the proof is sort of in the pudding. With physics, even the experts can go on developing theory which (as I understand it) arguably might never be testable, or is only testable after some absurdly expensive new machine is planned and built which might take decades. So an armchair physicist with an idea can have no route to falsifying their idea.
- tomcat27 5y ago
  sounds like a good plan to get tenure. lol

cinntaile 5y ago

The references should allow you to follow some sort of red thread. They will probably aid in closing the knowledge gap as well. It's probably like in this article (it was a good article btw) though, you need more maths. Either you use university curricula to find out what maths you need as a starting point or you focus on the maths that pops up in the papers and work your way from there.

tomcat27 5y ago
I can't speak for others, but I find scientific literature is not written general audience to understand, for a variety of reasons such as (a) publisher limitation on number of pages (b) to communicate core message right without distraction (c) to not bore a certain people etc. Not to mention, depending on the sub-fields, at least in AI, there is quite a bit of noise in papers as they are explaining _their_ viewpoint which might differ from broad consensus or disproven in future. I find papers are typically written keeping in mind a specific set of people, who are typically in that particular sub-field for some time. In sub-fields like quantum computation, that specific set of people authors have in mind while writing paper are <100, and those are not newbies. Newbies can quickly get discouraged reading a paper out of the blue. Picking the right papers at the beginning requires exposure to the field. It's not simple. I really hope what you said is practical. I tried. Hope it's my fault because the alternative option doesn't exist for me.

bmitc 5y ago

I think part of the problem is that science is not honest with itself. There is some sense that getting a Ph.D. is an institutionalization process, one that does not really reward non-conservative, long-term, or high probability of failure work. And god forbid if someone wants to revisit "known" ideas and results to help clarify them, apply them, or place them in better context. There is very much an "in" crowd in science, which leaves a lot of people, people who could positively contribute, feeling "out". We have commoditized academic research, at least in the science and engineering fields, which makes things very streamlined and on rails. I've spent some time in both academic and industrial circles of mathematicians, engineers, computer scientists, and physicists. It is a bit astonishing just how people in each one of these groups all sound like each other and distinct from the other groups. There are particular mannerisms and even idiomatic phrases that each group uses. Some simple ones are that physicists like to say "it goes like" and use the word "codes" to refer to code, as in software code. Computer scientists like to say "it better be". There are many others that I've lost track of and should have written down over these years. Yes, in many ways, the little idiomatic phrases make sense, but it at least lends credence to the question that if they all use the same phrases, something very simple and harmless, how else are they "conditioned"? The book Disciplined Minds: A Critical Look at Salaried Professionals and the Soul-battering System That Shapes Their Lives by Jeff Schmidt covers some of this. It all makes me wonder if someone like John Archibald Wheeler, a personal favorite, could be successful in today's academic climate.

nautilius 5y ago
The simple answer is: funding. There is no (public) money to revisit "known" ideas and results to help clarify them, apply them, or place them in better context. And I would not be the one to explain to the public to spend their money on known ideas rather than discover something new and shiny. Concerning a PhD, I think it's pretty easy to see that that's not the stage in your career where you would want to embark on long-term, high risk research. Also, I am not surprised that people doing a similar job have a common vocabulary and vernacular. You could say culture. I'd be surprised if carpenters, mechanics, et al. don't have the same thing, with other words and topics, of course. No conspiracies needed. Wheeler had significant contributions, had the right pedigree, the right schools. I don't see how he would be at a disadvantage today.
- bmitc 5y ago
  It's not about conspiracy theories. It's about reality. You even explicitly called it out: culture. That's just the way humans work, for better or worse. One of my points, and I think a point of Disciplined Minds, is that scientists are sometimes not honest or self-aware of the culture that they're a part of. Science is not a purely rational endeavor. There's some fashion and belief to it. > Wheeler had significant contributions, had the right pedigree, the right schools. I don't see how he would be at a disadvantage today. I didn't mean it as a critique upon Wheeler's potential. He was incredible. But he switched fields and research directions several times during his long and successful career. I'm not sure academics today have that fluidity in their careers (not due to internal forces but rather the external ones).
  - nautilius 5y ago
    I don't see why scientists would be less self-aware of their culture than other groups. In particular, that culture has got nothing to do with the things you decry (why is no one looking at old ideas, where, again, the simple problem is money). As soon as Wheeler would have tenure today, he could switch all day long. I am not talking about Wheeler's potential either, but whether he ticks the boxes to make him likely to succeed today. And he does. You are looking for a career that wouldn't work today? Freeman Dyson. Getting tenure at Princeton aged 29 without a doctorate, not so likely today.

car_analogy 5y ago

> I'm probably in the same boat as those people, but in AI. Some of the problems I find interesting are just out of my depth. You are not. There's an ocean of difference between someone that recognizes their ignorance, and someone that doesn't.

c22 5y ago
Surely if someone is paying a consultant to weigh in on their ideas they must have some inkling that they're out of their element. From the article: > One or two seemed miffed that I didn’t immediately exclaim: ‘Genius!’, but most of my callers realised that they can’t contribute to a field without meeting today’s quality standard. Then again, I hear only from those willing to invest in advancing their education to begin with.

abeppu 5y ago

But ... I think with AI, if you have an insight that isn't motivated by math per se, but is motivated by some other problem intuition, you can still attempt to implement it, and the proof is sort of in the pudding. With physics, even the experts can go on developing theory which (as I understand it) arguably might never be testable, or is only testable after some absurdly expensive new machine is planned and built which might take decades. So an armchair physicist with an idea can have no route to falsifying their idea.

tomcat27 5y ago
sounds like a good plan to get tenure. lol