rckoepke

42 ms·

361.
by rckoepke 7y ago
People who predict #2 ("sharp reversal of software developer salaries") are modeling software development jobs as an "inelastic" commodity. Oil is the quintessential inelastic commodity, so I can break that one down as a
362.
by rckoepke 7y ago
In all honesty I've felt that Google's Cloud Platform has been delivering really great product during Thomas Kurian's time. Not sure if that's actually related to him being there, but I've been super impressed with
363.
by rckoepke 7y ago
Technically you can do it in any language, but in software engineering we tend to stand on the shoulders of giants in order to get the job done on time. A lot of original excellent data processing, statistical analysis, and ML libraries wer
364.
by rckoepke 7y ago
You can do pose estimation with basic statistics?
365.
by rckoepke 7y ago
Fairly certain that's a feature, not a bug. (Obviously to hide other features which the end users would take offense to)
366.
by rckoepke 7y ago
For a year or two I had a mantra: "A day is a length of time that no man is rich enough to waste". It wasn't grammatically/syntactically valid, but it worked well as a token for a concept that helped me immensely those y
367.
by rckoepke 7y ago
Probably something very concise/dense/terse. https://blog.revolutionanalytics.com/2012/11/which-programmi... Potentially J or Lua, which were not analyzed in that article. However, Python has a lot of di
368.
by rckoepke 7y ago
> study groups, structure, people to talk to and discuss with and even just "we meet every Thursday at noon to ..." are not negligible. Indeed. These resources are stupendously valuable, and probably somewhat easier to find&#x2
369.
by rckoepke 7y ago
LeelaChess Zero could be considered an attempt at reproducing AlphaZero. GPT-2 (fully weighted) was arguably reproduced as well, eventually 'forcing' OpenAI to release the model they had held back [0]. 0: https://www.re
370.
by rckoepke 7y ago
Looks like overview to me. Guys like Francois Chollet aren't going to spend their time doing an in-depth tutorial, and even if he spent the whole 90-minute lecture walking you through the specific thing he invented (Keras), it probably
371.
by rckoepke 7y ago
And honestly those commodity AI networks from India will probably solve 90% of corporate needs, when properly pipelined.
372.
by rckoepke 7y ago
I believe Google's AutoML is attempting to answer these types of questions. It's obviously internal-only so others can't fork the research...but it has helped them invent new specific networks like "EfficientNet for Edge
373.
by rckoepke 7y ago
This is fantastic, but would really love if I got realtime feedback on my main pitch as I am making it, instead of the full spectrogram 1-5 seconds later. Like a little line chart.