31 lines
1.3 KiB
Org Mode
31 lines
1.3 KiB
Org Mode
#+TITLE: Untitled rant on machine learning hype
|
|
#+AUTHOR: Chris Hodapp
|
|
#+DATE: February 24, 2018
|
|
#+TAGS: technobabble
|
|
|
|
The present state in machine learning feels like an arms-race for
|
|
techniques that perfomr better, faster, more efficient, or whatever on
|
|
a handful of problems, and not much in terms of killer applications
|
|
that actually need this.
|
|
|
|
We've all been hearing for a few years about the demand here but
|
|
mostly there seems a dearth of companies that actually have any sort
|
|
of sustained vision for actual uses of machine learning. Plenty exist
|
|
that have grand promises, and plenty of large companies keep trying to
|
|
acquire all talent to further that arms race, but that's about it.
|
|
|
|
Certainly this will change as machine learning "gets better" but in
|
|
order for a lot of improvement to occur there must be at the same time
|
|
some actual compelling ideas and applications to drive it.
|
|
|
|
In that sense I don't believe that crrent advancements will be that
|
|
fruitful on their own. We don't need optimizations, we need
|
|
applications.
|
|
|
|
Of course this is not the first time a entire industry was imagined
|
|
and hyped based on neat technology and little else...
|
|
|
|
Right now I feel as though the work is going to those who can actually
|
|
articulate the "why" in specific terms, not those with some good
|
|
knowledge primarily on the "how".
|