is machine learning overhyped?

L3 Technologies / Eng chHh13
Nov 2 22 Comments

seems like everyone wants to go into Al and build hotdog not hotdog models. But i don’t understand the hype. ML seems to just be writing regressions and tuning your parameters. Any ML engineers out there wanna shed some light on this?

comments

Want to comment? LOG IN or SIGN UP
TOP 22 Comments
  • Intel / HR Meeoww
    Every technology is overhyped when new. Look at Hadoop from 10 years ago, Openstack 7 years ago. It's Kubernetes and Machine Learning now .
    Nov 2 0
  • Amazon / Eng tylium
    In my experience it rarely adds value. Usually a simple statistical model suffices.

    The few applications that are practical are speech recognition, image recognition, and self driving algorithms.

    For your run of the mill app/website you probably don't need it. Even for things like product recommendations.
    Nov 2 3
    • Zillow Group gentle
      Depends what you define ml. Those statistical models you mentioned are under ml umbrella.
      Nov 2
    • Amazon / Eng tylium
      True. In my definition applying a normal distribution or using collaborative filtering is just stats. While neural net is ML.
      Nov 2
    • Zillow Group gentle
      Your definition or mine, there are of opportunities to optimize different processes using data driven approaches (ml)
      Nov 2
  • Microsoft 3jrie44
    It's not hype. It's real.
    Nov 2 0
  • Zillow Group gentle
    Disclaimer: I'm doing ml so I'm biased. I'd say it's real so far at least. There are so many new usecases that are possible now that weren't a few years ago.
    Nov 2 5
    • Axtria 🐨 koala
      Are you an MLE or DS or AS? What's was your interview process like at Zillow?
      Nov 2
    • Zillow Group gentle
      AS. Our team does an exercise where you need to solve a data driven problem and write a report. Then phone screen is needed. Then onsite.
      Nov 2
    • Axtria 🐨 koala
      Can I pm you? I am in the process.
      Nov 2
    • Zillow Group gentle
      Sure
      Nov 3
    • Axtria 🐨 koala
      Thanks. DMed you.
      Nov 3
  • Microsoft CeAfu78
    ML is underrated.
    Nov 2 0
  • New tnTk52
    Depends on which part of ML. Some methods are widely used in practice and their impact has been properly understood rather than exaggerated. I found Gartner has a hype curve for ML, which I think is very informative. Here is the link: https://www.gartner.com/smarterwithgartner/top-trends-on-the-gartner-hype-cycle-for-artificial-intelligence-2019
    Nov 2 0
  • Yahoo / Eng hDgiT
    There are so many ML clowns in the industry.
    Nov 3 0
  • Salesforce moussaka🥗
    It's old stats and old math that started to be monetized in the last 25 years. More so in the last 10 owing to compute and distributed systems advancements.
    But it's not hype. The potential is real. The concepts are age old
    Nov 3 0
  • Google / Eng manoi3
    It's real
    Nov 2 0
  • Lyft wXfl07
    It's a 60-year-old tech and they think that having two layer neural networks makes it brand-new yesterday. if you look at AI classes they spend most of their time taking credit for work in other fields outside of AI.

    This is a fad that we have to endure every 25 years and mark my words it will not last!
    Nov 2 1
    • New ovGu51
      Didn't think Lyft would say this 🤔
      Nov 3
  • Google b6e
    It’s usually real. Complex analysis is not common in ML.
    Nov 3 0
  • ML is everywhere and it adds huge value. As you can tell already many repetitive jobs will be automated by it.
    Nov 3 0
  • Amazon dubeous
    POOP
    Nov 2 0

Salary
Comparison

    Real time salary information from verified employees