Google's morale crisis is about to get worse | Hacker News Reader

Google's morale crisis is about to get worse

theverge.com

21 pointsukuina19 comments

Text-only, no Javascript; works where archive.is is blocked

     #!/bin/sh
     tr @ '\34' \
     |sed 's/articleBody/@/;s/}/@/g' \
     |tr @ '\12' \
     |sed -n 's/\\n/@/g;s/&#8217;/'"'"'/g;
     s/&#8212;/-/g;s/&#8230;/.../g;
     s/&#822[01];/\"/g;s/\"$//;s/^\":\"//p' \
     |tr -d '\240' \
     |tr @ '\12' \
     |tr '\34' @

     curl -si40A "" https://www.theverge.com/2024/3/3/24089843/google-morale-crisis-about-to-get-worse|1.sh > 1.txt
     firefox ./1.txt (overkill)
     less 1.txt

I like that you think Firefox is the overkill in all of that. At least, your username compliments this post quite well.

I would also like to thank you for sending me down a pointless rabbit hole which led to this cleaner madness:

  curl -sL https://www.theverge.com/2024/3/3/24089843/google-morale-crisis-about-to-get-worse | grep -Pio 'application/ld\+json">\K[^<]+' | jq -r '.articleBody'

FWIW

I only use curl and firefox in HN examples. I do not use these programs myself.

sed and tr are each small, fast and ubiquitous. Found by default in GNU/Linux and BSD userland for the last 30 years. Both found in GNU/Linux and BSD multicall binaries.

I'm not a Perl RE user. Neither busybox nor NetBSD grep has a -P option.

Extracting text from HTMLJSON, as a one-off, on the computers I own (older, smaller, underpowered) might look something like

   echo https://www.theverge.com/2024/3/3/24089843/google-morale-crisis-about-to-get-worse \
   |yy025|nc -vv h6 80 > 1.htm

   yy059 < 1.htm|sed -n '/articleBody/{p;q;}'|yy032|less
   or
   yy059 < 1.htm|grep -m1 articleBody|yy032|less

h6 is a loopback-bound forward proxy

Static binaries

   81.6 KiB nc
   45.9 KiB yy025
   45.9 KiB yy032
   45.9 KiB yy059

By comparison

   234.2 KiB GNU grep (dynamically-linked to libpcre2)
   614.3 KiB libpcre2-8.so.0.11.2
   370.0 KiB libjq.so.1.0.4 (dynamically-linked to jq)
   607.6 KiB libonig.so.5.4.0 (dynamically-linked to libjq)

There are 37,291,836 SLOC in Firefox. Over 37 million lines of code. That's probably larger than the Linux kernel. The program requires hundreds of megabytes of disk space. It is overkill to use such a program to read a small text file. Are there faster alternatives that require less storage space, memory and CPU.

https://4e6.github.io/firefox-lang-stats/

All while Pichai rakes in $260M/yr for running Google into the ground and failing to retain top talent. As a Xoogler, I know for a fact most googlers wouldn’t be too miffed about slightly lower raise every once in a while, but not when the CEO and SVPs avail themselves of such lavish, and ever increasing, compensation packages

I'll take a WAG that one cause of these pay patterns is the infiltration of the MBA self-congratulation mentality into the ranks.

That and Google’s incredibly deep hierarchical structure. Nowhere else have I seen VPs report to VPs and directors report to directors. At Google this is quite common, and not just now - it was quite common under Schmidt and Page, too. It’s never clear what those people do, and by the time information filters up to the CEO (who, let me remind you, can’t tell a tensor from a hole in the ground) everything is hunky dory. Previously the CEOs were technical and much harder to bullshit. Sundar isn’t technical. He truly is the Ballmer of Google, probably setting the stage for someone else.

https://archive.is/7Qn5O

> While most employees will still be getting an overall increase to their compensation, the majority won’t be getting anywhere near as much of an increase as they have in years prior, according to internal emails and documents shared with managers that I’ve seen. For the “significant impact” rating that the majority of employees are assigned by their managers, there will be no salary increase in most cases.

The first and last sentence here seem to be at odds. Am I missing something? GSUs?

Seems like base salary is stagnating (i.e., no raises), but stock refreshers continue at last-year's levels?

Stocks as pay could be seen as the private company version of printing money, right?

See the issue with AI or (AGI) is that it is trying to model or imitate a system which is irrational, whimsical and inconsistent i.e. human. So the o/p of an AI system will be subjected to bias by humans especially when there is no universal right & wrong.

Can we not post links which are behind paywalls? Or instead link to some other source so we can read the article in question?

> Please submit the original source. If a post reports on something found on another site, submit the latter.

HN guidelines[1]

[1] https://news.ycombinator.com/newsguidelines.html

This feels like a challenge moving with the times. I strongly agree with the guidelines – paywalls are becoming pervasive entirely due to people not compensating journalists for their work – but the number of different subscriptions needed to follow more than a handful of writers is starting to cost as much as a cable TV plan.

Previously I thought that the subscriber share mechanisms would offer a bit of a balance here but with the entire industry worried about AI training I’m not sure that’ll help since they’re going to be worried about even a single retrieval not gated behind a login system where you agree to pay them more for training.

Is it possible we're reaching a world where we get our news from AI bots exclusively because the companies behind said AI bots are the only ones willing to pay for access to "The News (TM)" ?

It won’t be that simple but it’ll have some nasty effects: for example, media sites which are funded to support political viewpoints with a lot of money behind them are less likely to have paywalls because they can rely on sponsorship. That’s likely to make the AI summaries even less trustworthy because training will only be so successful at correcting for the selection bias of training data which is freely available or cheaply licensed. Most people know to some extent that this needs to be controlled for on really big issues like climate change but there are tons of smaller issues where someone working on an ML system won’t have any idea of the scale of the problem because they don’t know about issues which are largely only known within a particular regional, business, linguistic, etc. community.

What's the point in submitting a source which cannot be read? It makes no sense at all. If people can't read it, don't post it.

"We haven't fully understood why the model leans left" sure sounds a lot like "Reality has a well-known liberal bias".

Life imitating art, I suppose.

Reply on news.ycombinator.com