machine learning hype in FANG?

Magic Leap iugder24
Oct 28 22 Comments

Had an onsite interview at FB for a general engineering role.

EVERY single interviewer said they were working on machine learning related projects. Is that what it is now in FANG that everybody is wanting to somehow get involved in ML?

123 VOTES SELECT ONLY ONE ANSWER
VOTE VIEW RESULT

comments

Want to comment? LOG IN or SIGN UP
TOP 22 Comments
  • NVIDIA acfm
    ML is a buzzword. I’ve seen people who claim their old ordinary least squares is suddenly ML
    Oct 28 2
    • Google djkn deg
      The question is why do you think ols is not machine learning?
      Oct 28
    • Fannie Mae sDEB48
      It totally is machine learning, all regression is machine learning. You created a predictor using sample data. Fitting a line with gradient descent instead of least squares is a one layer neural net with no nonlinearity.

      I think the fact that people think it’s funny to call it machine learning is the result of ML being overhyped.
      Oct 28
  • Amazon syrbskcb
    ML is the 'big thing' right now. It used to be big data and mobile development a few years ago. I expect 'quantum computing' to pick up in a few years.
    Oct 28 3
    • Google pissoff
      Too far away
      Oct 28
    • Google / Eng manoi3
      Maybe a decade or so. This is just like when Google did self driving cars a decade back.
      Oct 28
    • Amazon syrbskcb
      Lvl 5 Teslas by 2020 - Elon Musk
      Oct 28
  • Google pissoff
    Not sure which product from Facebook actually uses ML. Their recommendations for me usually sucks. Recommendation systems are not ML. It could be easy done just by using click tracking.
    Oct 28 9
    • NVIDIA acfm
      They have/had a whole applied machine learning org exactly to apply ML to business problems
      Oct 28
    • Google pissoff
      Like what?
      Oct 28
    • Waymo PdAX50
      Click tracking usually is machine learning. What do you regress clicks against?
      Oct 28
    • Google pissoff
      Click attribution models are not machine learning models. There's usually no feedback loop. Last click, first clicks etc are well defined models. They don't evolve over time.
      Oct 28
    • Waymo PdAX50
      Hmm, maybe it varies by PA. When I worked at G, in my PA, that was not the case.
      Oct 29
  • Cisco Owl 🦉
    I just wish FB would make timeline filtered by friends and followed pages and order by the time posted. I hate seeing friends’ new contents like 5 days later.
    Oct 28 0
  • Google
    MacDouble

    Google

    PRE
    Google
    MacDoublemore
    Kid do a linear regression and say I created a machine learning model to do xyz...
    Oct 28 2
    • Google / Eng manoi3
      Kid use angular and claims making a single page webapp going abc
      Oct 28
    • Google pissoff
      That's fine in my opinion. You cultivate corn and say you produce corn. But then you cultivate corn and say you created an energy source for humans. That's just glorification.
      Oct 28
  • New owLL68
    It's not hype for FB and Google to do ML as they have tons of data to train their models. Where there are repetitive patterns, there are potentials to use ML. Of course, you can solve some problems somehow without using ML. But those methods are usually highly ad-hoc which involve tons of tricks and magical intuition, it's hard to generalize to other similar problems. Moreover, your tons of tricks can easily fail to achieve any reasonable results with high dimensional problems.
    Oct 29 0

Salary
Comparison

    Real time salary information from verified employees