ta_tunestub

LeRobot SO-101

Not just recruiters, but hiring managers.

I trained the model used to rate candidates’ resumes at a largish public company (based on hiring manager feedback signals), and when you inspected resumes with “CEO” as a job title the expected scores (all else being equal) were much lower.

“The Book of Why” and “Statistical Rethinking” course [1]

[1]https://twitter.com/chrismgreer/status/1714687870286655885?s...

Yep, there are many reimplementations. Here is a reimplementation that swaps out a neural net with a GBDT to address compute costs:

https://github.com/cgreer/alpha-zero-boosted

Are you aware of any code/projects that convert something like a fully trained resnet50 model into a maddness optimized, approximate model?

An approximate but speedier resnet inference model that runs on a CPU would be useful even if it’s not quite as fast/accurate as a GPU inference model, since currently the cost to run a GPU is typically higher than CPUs.

Amen!

According to Census Bureau (2019)[1], this is the racial/ethnic makeup of the US:

    White                                       60.4%
    Hispanic and Latino Americans (of any race) 18.3%
    Black or African American                   13.4%
    Asian                                       5.9%
    Native Americans and Alaska Natives         1.3%
    Native Hawaiians and Other Pacific Islander 0.2%
    Two or more races 2.7%

Below is the normalized 2019 CS Ph.D. count for what you'd expect if the US was 100% of each respective race/ethnicity:

    White                          608
    Hispanic...                    115
    Black or African/American      97
    Asian                          2,560
    Native Americans/Alaska Native 153
    Native Hawaiian/Pac Islander   500

[1] https://en.wikipedia.org/wiki/Race_and_ethnicity_in_the_Unit...

> which of these "versions" of the project would be the easiest for me to understand/implement?

I have the same question. Not sure I have an answer yet, but this paper includes some pseudocode that implements the algorithm: https://arxiv.org/src/1911.08265v1/anc/pseudocode.py

I'm planning on trying to train something simple like TicTacToe to both see if it works and understand how it works.

FWIW, This same type of event DID demotivate me and a lot of my co-workers who were early employees at a startup (now 1000+ person company). The company raised a round and the founders sold off some of their shares. We knew because they bought nice cars/houses after the event. Our shares got diluted.

From our perspective, it seemed like a slight vote of no confidence in the company and a little unfair to us that we didn't get the chance to sell shares whereas they did. You lose a bit of that "we're in this together" vibe that makes it organically motivating to work at a place.

Just my 2c, but consider arranging a situation where employees can sell up to a percentage of their options as well. A small percentage makes the eventual payoff palpable and will reduce resentment. You'll likely need to sustain the high levels of motivation for your fast-growing startup to continue it's success.