7 ms·

One of the “minimum requirements” “ 8+ years of experience in one or more of the following areas: machine learning, recommendation systems, pattern recognition

One of the “minimum requirements” “ 8+ years of experience in one or more of the following areas: machine learning, recommendation systems, pattern recognition, NLP, data mining, or artificial intelligence”

yangminded 5y ago
That's entirely feasible. Mind that 8 years ago was 2013 and systems like the recommendation system from Amazon was already online. Also it only specifies machine learning, pattern recognition, NLP etc - all of which are active research fields for far more then 20 years already. It's not like they are asking for more than 8 years of experience in Pytorch.
- user-the-name 5y ago
  Sure, it's feasible on its own. But the position is for working on graphics rendering, a completely unrelated field. Are you supposed to have had two careers at the same time, working eight years in AI while also working all that same time on graphics?
  - mkr-hn 5y ago
    There's a lot of research on using AI/ML with graphics. I've seen demos people have done using it to create HD remasters of shows originally in SD on tape or too-small film, or where there wasn't much interest from IP owners in doing it. Or where, once you expanded sideways, there was too much crew or rough set to use it. Since Facebook is doing VR, and this position is about VR, I can easily see the potential of using ML to make the most of sketchy/low-quality models and textures people might want to use in worlds. Imagine if you could upload a rough model, describe what you want, and get it made for you.
    - seoaeu 5y ago
      Yeah, but back in 2013? Most of what you're talking about is much more recent. To have 8 years of AI/ML expertise already while also previously being a graphics expert requires a very improbable career progression
    - Scarblac 5y ago
      Yes, but _on mobile_? With details like "graphics hardware and drivers on mobile platforms" and "the mobile graphics pipeline and interaction across display and optics". I mean obviously a lot of people work in that area but the overlap with AI/ML simultaneously seems very small, for 8+ years in one person. And of course if you're in research then it's hard to tick boxes like "leading consumer-based projects with industry-wide impact" and "mentoring/influencing senior engineers across organizations" and so on.
      - TeMPOraL 5y ago
        [snip; not enough coffee]
        user-the-name 5y ago
        Try re-reading the post you responded to.
- Aerroon 5y ago
  But how many people are also going to be "highly experienced in the mobile graphics pipeline"? Isn't that also about ~10 years old?
  - marcinzm 5y ago
    When you're willing to pay a million+ a year you just need a couple people.
    - thrill 5y ago
      When you are willing to pay a million+ a year you already know who you want to hire.
- dalbasal 5y ago
  Feasible, but in a naive sense. much sk The number of people working in "those areas," defined narrowly, is much bigger than it was eight years ago. That makes this a big requirement. People who are skilled, deep or enthusiastic enough to make that 8 years experience better than 4 are not likely to just show up, while also ticking the other boxes. Narrowing the pool past a certain point is a different type of problem. In any case, these aren't to be taken to literally.
sbarre 5y ago
While I agree the overall minimum requirements for this job are a bit silly, most (all?) of those topics have been around for more than 8 years, in some shape or form.
- viraptor 5y ago
  For NLP, the famous Alice is from 1995 and ELIZA is from 1960s. You can easily have 8 years of experience in this area.
  - kevinmchugh 5y ago
    Eliza is like 8 if conditions. It's a parlor trick
    - viraptor 5y ago
      It doesn't matter how simple it was. For its time, that was the state of art as far as I can tell. There's lots of things invented decades ago that would be a parlor trick today.
- CrazyStat 5y ago
  Each requirement taken individually is plausible, but the intersection of them all is absurd.
  - sbarre 5y ago
    Agreed! Perhaps that's why the listing says "one or more" and not "all"
    - CrazyStat 5y ago
      It only says that about the various ML/AI requirements, which is a single bullet point. That doesn't help the intersection between bullet points, which is the issue.
- alistairSH 5y ago
pmontra 5y ago
It seems that the whole AI field was started four or five years ago but actually it's several decades old. A recent example: for how long did Amazon use recommendation systems? And data mining was called OLAP [1] in the 90s and probably with another name in the 60s. [1] https://en.wikipedia.org/wiki/Online_analytical_processing https://en.wikipedia.org/wiki/Online_analytical_processing
988747 5y ago
Professor Geoffrey Hinton would have 30+ years of experience with machine learning, I think.
- long_time_gone 5y ago
  But probably 0 years experience in "the mobile graphics pipeline and interaction across display and optics", which is the actual job.
- htrp 5y ago
  Can't hire... failed the leetcode test
dcolkitt 5y ago
IMO using pre-determined time windows to filter on experience is always a mistake. Smart people acquire mastery in much shorter chronological time. This is literally the definition of intelligence and it’s been repeatedly verified over decades by industrial psychology. If Linus Torvalds decided to pick up data engineering tomorrow, he’d be far better at it in 3 months than the average dev who’s been in the field for 10 years. When you filter on years of experience, you’re inadvertently biasing towards less intelligent candidates. Smart people with that much experience will have enough mastery that they’re not looking for a mid level position. Polymaths who jump between fields, quickly building up competence won’t meet your arbitrary threshold.
- ratww 5y ago
  > Smart people acquire mastery in much shorter chronological time ...and lots of people don't, even after 10 years working in the field. Not because they're not smart, of course, but it's common.
- tester34 5y ago
  >If Linus Torvalds decided to pick up data engineering tomorrow, he’d be far better at it in 3 months than the average dev who’s been in the field for 10 years. bold claim, I'd want to take a bet with you for real money on that decade ago was 2011, wasn't it time where applied data engineering started becoming cool? thus majority of people shouldnt be those, who started learing it due to $$, but out of curiosity? idk
  - test_epsilon 5y ago
    He kind of already did that -- designed and wrote git in a month.
    - tester34 5y ago
      I don't think I'd call git "data (science) engineering tool"
      - 5y ago
marcinzm 5y ago
Machine learning, data mining and pattern recognition aren't brand new fields in any meaning of the word. I've had classes on them 20 years ago in college and I know a lot of people with 8+ years of experience (including myself). Hell, my dad worked on them when he was in college 40+ years ago. Facebook pays more than enough to get world class experts so why would they not aim to hire them?

yangminded 5y ago

That's entirely feasible. Mind that 8 years ago was 2013 and systems like the recommendation system from Amazon was already online. Also it only specifies machine learning, pattern recognition, NLP etc - all of which are active research fields for far more then 20 years already. It's not like they are asking for more than 8 years of experience in Pytorch.

user-the-name 5y ago
Sure, it's feasible on its own. But the position is for working on graphics rendering, a completely unrelated field. Are you supposed to have had two careers at the same time, working eight years in AI while also working all that same time on graphics?
- mkr-hn 5y ago
  There's a lot of research on using AI/ML with graphics. I've seen demos people have done using it to create HD remasters of shows originally in SD on tape or too-small film, or where there wasn't much interest from IP owners in doing it. Or where, once you expanded sideways, there was too much crew or rough set to use it. Since Facebook is doing VR, and this position is about VR, I can easily see the potential of using ML to make the most of sketchy/low-quality models and textures people might want to use in worlds. Imagine if you could upload a rough model, describe what you want, and get it made for you.
  - seoaeu 5y ago
    Yeah, but back in 2013? Most of what you're talking about is much more recent. To have 8 years of AI/ML expertise already while also previously being a graphics expert requires a very improbable career progression
  - Scarblac 5y ago
    Yes, but _on mobile_? With details like "graphics hardware and drivers on mobile platforms" and "the mobile graphics pipeline and interaction across display and optics". I mean obviously a lot of people work in that area but the overlap with AI/ML simultaneously seems very small, for 8+ years in one person. And of course if you're in research then it's hard to tick boxes like "leading consumer-based projects with industry-wide impact" and "mentoring/influencing senior engineers across organizations" and so on.
    - TeMPOraL 5y ago
      [snip; not enough coffee]
      - user-the-name 5y ago
        Try re-reading the post you responded to.
Aerroon 5y ago
But how many people are also going to be "highly experienced in the mobile graphics pipeline"? Isn't that also about ~10 years old?
- marcinzm 5y ago
  When you're willing to pay a million+ a year you just need a couple people.
  - thrill 5y ago
    When you are willing to pay a million+ a year you already know who you want to hire.
dalbasal 5y ago
Feasible, but in a naive sense. much sk The number of people working in "those areas," defined narrowly, is much bigger than it was eight years ago. That makes this a big requirement. People who are skilled, deep or enthusiastic enough to make that 8 years experience better than 4 are not likely to just show up, while also ticking the other boxes. Narrowing the pool past a certain point is a different type of problem. In any case, these aren't to be taken to literally.

sbarre 5y ago

While I agree the overall minimum requirements for this job are a bit silly, most (all?) of those topics have been around for more than 8 years, in some shape or form.

viraptor 5y ago
For NLP, the famous Alice is from 1995 and ELIZA is from 1960s. You can easily have 8 years of experience in this area.
- kevinmchugh 5y ago
  Eliza is like 8 if conditions. It's a parlor trick
  - viraptor 5y ago
    It doesn't matter how simple it was. For its time, that was the state of art as far as I can tell. There's lots of things invented decades ago that would be a parlor trick today.
CrazyStat 5y ago
Each requirement taken individually is plausible, but the intersection of them all is absurd.
- sbarre 5y ago
  Agreed! Perhaps that's why the listing says "one or more" and not "all"
  - CrazyStat 5y ago
    It only says that about the various ML/AI requirements, which is a single bullet point. That doesn't help the intersection between bullet points, which is the issue.
alistairSH 5y ago

pmontra 5y ago

It seems that the whole AI field was started four or five years ago but actually it's several decades old. A recent example: for how long did Amazon use recommendation systems? And data mining was called OLAP [1] in the 90s and probably with another name in the 60s. [1] https://en.wikipedia.org/wiki/Online_analytical_processing https://en.wikipedia.org/wiki/Online_analytical_processing

988747 5y ago

Professor Geoffrey Hinton would have 30+ years of experience with machine learning, I think.

long_time_gone 5y ago
But probably 0 years experience in "the mobile graphics pipeline and interaction across display and optics", which is the actual job.
htrp 5y ago
Can't hire... failed the leetcode test

dcolkitt 5y ago

IMO using pre-determined time windows to filter on experience is always a mistake. Smart people acquire mastery in much shorter chronological time. This is literally the definition of intelligence and it’s been repeatedly verified over decades by industrial psychology. If Linus Torvalds decided to pick up data engineering tomorrow, he’d be far better at it in 3 months than the average dev who’s been in the field for 10 years. When you filter on years of experience, you’re inadvertently biasing towards less intelligent candidates. Smart people with that much experience will have enough mastery that they’re not looking for a mid level position. Polymaths who jump between fields, quickly building up competence won’t meet your arbitrary threshold.

ratww 5y ago
> Smart people acquire mastery in much shorter chronological time ...and lots of people don't, even after 10 years working in the field. Not because they're not smart, of course, but it's common.
tester34 5y ago
>If Linus Torvalds decided to pick up data engineering tomorrow, he’d be far better at it in 3 months than the average dev who’s been in the field for 10 years. bold claim, I'd want to take a bet with you for real money on that decade ago was 2011, wasn't it time where applied data engineering started becoming cool? thus majority of people shouldnt be those, who started learing it due to $$, but out of curiosity? idk
- test_epsilon 5y ago
  He kind of already did that -- designed and wrote git in a month.
  - tester34 5y ago
    I don't think I'd call git "data (science) engineering tool"
    - 5y ago

marcinzm 5y ago

Machine learning, data mining and pattern recognition aren't brand new fields in any meaning of the word. I've had classes on them 20 years ago in college and I know a lot of people with 8+ years of experience (including myself). Hell, my dad worked on them when he was in college 40+ years ago. Facebook pays more than enough to get world class experts so why would they not aim to hire them?