The thing I learned from Amazon's senior principals is that actually it's good and normal to turn red in the face and scream at your junior colleagues that they're fucking idiots when they have the temerity to politely disagree with you.
They get it from senior management, and pass it down like generational trauma. This was a problem even in 2013 when I worked there. Once, I was new and actually pushed back against a Director level person's poor behavior in a 70 person meeting, because I didn't know better. I was approached by multiple individuals afterwards telling me how "brave" it was of me.
At Amazon, unkind and downright unprofessional behavior by people higher up the chain is normalized, and has been for a very long time.
I've never worked directly for Amazon, but for a consultancy that was an AWS Partner.
I got an invite to a team skip level meeting once, and holy shit I could not believe the asshole and bullshit crap those seniors were tossing at each other, at the Partner manager, and also us.
"Strive to be Earth's Best Employer" coming from Amazon leadership is maybe best understood as a joke. Amazon is a meat grinder, even more so since the big tech layoffs started.
(This kind of comment always elicits current Amazon engineers who disagree because they haven't personally experienced this. To them, I say: Stay at Amazon long enough and it /will/ happen to you. To those currently in the grinder: I hit the eject button at L6 and found a much better gig; it gets better!)
I had lunch with the Amazon leader most responsible for ensuring all staff in the fast-moving-cardboard half of the company had health insurance from day one of employment, no waiting time. Of a decades long career, that was the one thing I saw most animate her—care for fellow humans.
When the 90th percentile employee has a GED and works warehouse or delivery, actions to earn “best employer” may be invisible or worse to the 5% who are software people. I’ve worked at Meta too, and Meta absolutely had better coffee. And WAY better health insurance. But Amazon’s health insurance is uniform for all staff, and that means something.
In Seattle there are some developers who call Amazon employees AmHoles. It's part their general arrogance, punctuated by things like their tendency to walk down the sidewalk four abreast and not notice that they are pushing people going the other way out of their way.
I worked a short contract there and I've seen the 'meatgrinder' bit. I joked (not really joking) with my fellow contractors that maybe the reason they walk four abreast is shell-shock, not arrogance. A couple days a week we went to lunch in a daze.
It's clear there's not enough quality control in their 'culture', by almost an order of magnitude. I've known two different people who quit after less than 2 weeks. One after being called on Sunday asking why he wasn't at work. On his second fucking week at the company.
Kind of amazing the author just takes everything at face value and doesn't even consider the possibility that there's a hidden layer of instructions. Elon likes to meddle with Grok whenever the mood strikes him, leading to Grok's sudden interest in Nazi topics such as South African "white genocide" and calling itself MechaHitler. Pretty sure that stuff is not in the instructions Grok will tell the user about.
The "MechaHitler" things is particularly obvious in my opinion, it aligns so closely to Musk's weird trying-to-be-funny thing that he does.
There's basically no way an LLM would come up with a name for itself that it consistently uses unless it's extensively referred to by that name in the training data (which is almost definitely not the case here for public data since I doubt anyone on Earth has ever referred to Grok as "MechaHitler" prior to now) or it's added in some kind of extra system prompt. The name seems very obviously intentional.
Grok was just repeating and expanding on things. Someone either said MechaHitler or mentioned Wolfenstein. If Grok searches Yandex and X, he's going to get quite a lot of crazy ideas. Someone tricked him with a fake article of a woman with a Jewish name saying bad things about flood victims.
> Pretty sure that stuff is not in the instructions Grok will tell the user about.
There is the original prompt, which is normally hidden as it gives you clues on how to make it do things the owners don't want.
Then there is the chain of thought/thinking/whatever you call it, where you can see what its trying to do. That is typically on display, like it is here.
so sure, the prompts are fiddled with all the time, and I'm sure there is an explicit prompt that says "use this tool to make sure you align your responses to what elon musk says" or some shit.
Of course they are. The interesting thing isn't how good LLMs are today, it's their astonishing rate of improvement. LLMs are a lot better than they were a year ago, and light years ahead of where they were two years ago. Where will they be in five years?
What's to stop ICE from declaring anybody they don't like an alien and rendering them to the Salvadoran prison? We already know they're inventing false pretenses to classify people as gang members. There is no functional oversight of what they're doing.
> Even the most egregious cases, which are very bad, have left a paper trail for lawyers and journalists to follow.
We don't know the identities of, or even how many people have been extrajudicially rendered to the Salvadoran prison. The administration claims they're not citizens, but how would we know?
