Yes the solution is pretty simple here. You have no obligation to work weekends unless you're getting some direct monetary benefit from it.
That's not how you create great things anyway, that's just what naive 20yr olds think hard work looks like. Rushing things, not taking breaks, or ever disconnecting your brain vs creating a consistent high output rhythm over a long period by getting into the zone, and only slamming for legit emergencies.
> You have no obligation to work weekends unless you're getting some direct monetary benefit from it.
That's not how you create great things anyway
It shouldn’t be systematically. But when I did a start-up, this was an opening that let us displace—in one case, put out of business—the competition: identifying the moments where working unconventionally creates asymmetric value.
Also, we’re talking about AI engineers. If you’re being paid seven figures and refusing to ever work weekends you should seek out a lifestyle role. (And to be clear, I have a tonne of respect for people who make that choice. You’re just not going to command a multi-million dollar comp package.)
This is why Kanban continues to be the only development methodology that I believe in.
"Let's go look at the board. Which of these other high priority items in flight would you like to deprioritize? By the way, our back of the envelop discovery session shows this requested feature to be six tasks, which historical TEAM velocity indicates will be done much later than your requested deadline. Also our metrics for other past weekend heroics indicate that we're going to see a 20-50% decline in productivity for the next two weeks, so we should expect the following work to be late."
At least in my experience it's the best way to make stakeholders stare at the reality of what they're asking.
Of course, its also a great indicator for people who aren't willing to look at reality. I don't know what the solution to that is, but there are a lot of reasonable people out there when you find a good way for them to understand the forces involved.
A single person saying “no” just means they will find the next person to do it. In return, person that said “no” will face consequences. Impossible to complete work. Slightly more hostile work environment. “Evidence” the management will use to create PIP. PIP will probably be work that cannot be completed within timeline.
Shit like this wouldn’t fly in a union.
Personally, I wouldn’t give 2 shits. I would love to see the face of the douchebag that asked me this and I just left that person on unread.
In this market, AI engineers get hired easily. With recent ruling on non-compete clauses, this makes it much more easier to move between firms.
Everyone is really attached to having the AI equivalent of the "iPhone Moment", but I think we already had it. It was just spread out over multiple products, from multiple companies, over the past few years.
There's obviously still enthusiasm for machine learning things, but you've definitely lost the "wow" factor for most people. Better to actually build the iPhone 3GS first, as it were.
From my perspective, this is AI bubble part 3 or 4. DARPA and UKI AI funding for chatbots, perceptrons dried up in the 70s; expert systems and symbolic reasoning dried up in the late ‘80s and ‘90s. Through the ’00s and ‘10s, ML/DL was successful, in part by shedding the guise of AI. Now we’re in an unsupervised learning and large-scale multilayer transformer networks wave, and while I expect the financing bubble for these projects will pop sooner than later, the tech will continue to live on and be integrated into future projects, as have all other previous AI breakthroughs.
This happens with the big 3 telecom companies as well. Too much competition leads to people spending too much overtime trying to compete with each other, because if one adds a feature, everyone else needs to put out that fire.
That is why I like movements like Apple, who don't compete. They simply add a feature that another competitor has to their to do list and set up a game plan for 2-3 years to implement.
I also like this about Google whose products are so massive that in order to make substantial improvements, they would need to do an entire restructuring for the long game.
both of these initiatives are the long-game plays. And so is the effort. why can't the same be happening in stuff like telecom, and AI? Don't worry about small time attrition of users. Worry about the long game.
Generally AI engineers are hugely compensated, compared to overall massively compensated IT folks. In such such cases I struggle to find much sympathy, in same vein when some wall street guy complains about burnout while driving some V12 Ferrari from some old paintings auction.
There is always the option to just quit and ask for maybe 50% more at next company struggling desperately with same stuff.
Keep in mind that this article isn't necessarily talking about the folks who are making the models.
There is a ton of much more ordinary software engineering work that goes into building this stuff. The people being talked about in this article could just be frontend or backend engineers who were working on a product listing page this time last year.
does it matter if they are hugely compensated or not? I don’t understand this. General agreement is usually set when you sign on. Anything outside this agreement would be met with a stern “no”. Now if company wants to airdrop an extra amount into my bank account for this adhoc work, we can discuss terms. Otherwise, …
“Fuck you, pay me”
Not sure why this is so hard. Makes it even difficult when you have bendy co-workers.
Generally, this is not always true. Generally, you’re being a bit of a jerk.
You are basically saying "Pffft. First world problems." Which is always a flippant dismissal of someone's actual problem, and you're doing it in HN of all places. A website by YCombinator, which is more or less the tip of the spear of the first world.
Stories like this one prove startups have a real chance in the space because they are not waylaid by the politics and conflicting incentives of various orgs inside of the large tech companies
AI is making even more established companies behave like early startups. shifting deadlines, building things to appease investors and fight for funding, learn how to fly the plane while building the engine, having to set your own boundaries, etc.
It’s always strange for me to hear anecdotes from “not-early-startup” folks complaining about not knowing how to set boundaries or needing learning to be spoonfed to them when they’re exposed to an early-startup-like environment.
I always wonder if they’re in the majority or I am.
tldr; too much capital necessary for wide participation, secrecy and paranoia is a key part of building a moat, and questionable training data. the high salaries just add to the pressure.
Readit News
Why don't people feel empowered to say no to this sort of thing? Is the market for engineers so bad that we have to tolerate these things out of fear?
That's not how you create great things anyway, that's just what naive 20yr olds think hard work looks like. Rushing things, not taking breaks, or ever disconnecting your brain vs creating a consistent high output rhythm over a long period by getting into the zone, and only slamming for legit emergencies.
Tech needs to be unionized.
It shouldn’t be systematically. But when I did a start-up, this was an opening that let us displace—in one case, put out of business—the competition: identifying the moments where working unconventionally creates asymmetric value.
Also, we’re talking about AI engineers. If you’re being paid seven figures and refusing to ever work weekends you should seek out a lifestyle role. (And to be clear, I have a tonne of respect for people who make that choice. You’re just not going to command a multi-million dollar comp package.)
"Let's go look at the board. Which of these other high priority items in flight would you like to deprioritize? By the way, our back of the envelop discovery session shows this requested feature to be six tasks, which historical TEAM velocity indicates will be done much later than your requested deadline. Also our metrics for other past weekend heroics indicate that we're going to see a 20-50% decline in productivity for the next two weeks, so we should expect the following work to be late."
At least in my experience it's the best way to make stakeholders stare at the reality of what they're asking.
Of course, its also a great indicator for people who aren't willing to look at reality. I don't know what the solution to that is, but there are a lot of reasonable people out there when you find a good way for them to understand the forces involved.
A single person saying “no” just means they will find the next person to do it. In return, person that said “no” will face consequences. Impossible to complete work. Slightly more hostile work environment. “Evidence” the management will use to create PIP. PIP will probably be work that cannot be completed within timeline.
Shit like this wouldn’t fly in a union.
Personally, I wouldn’t give 2 shits. I would love to see the face of the douchebag that asked me this and I just left that person on unread.
In this market, AI engineers get hired easily. With recent ruling on non-compete clauses, this makes it much more easier to move between firms.
There's obviously still enthusiasm for machine learning things, but you've definitely lost the "wow" factor for most people. Better to actually build the iPhone 3GS first, as it were.
Maybe once part 4 or 5 roll around it will be “stable”
Damn son of a bitch is ghosting me.
That is why I like movements like Apple, who don't compete. They simply add a feature that another competitor has to their to do list and set up a game plan for 2-3 years to implement.
I also like this about Google whose products are so massive that in order to make substantial improvements, they would need to do an entire restructuring for the long game.
both of these initiatives are the long-game plays. And so is the effort. why can't the same be happening in stuff like telecom, and AI? Don't worry about small time attrition of users. Worry about the long game.
There is always the option to just quit and ask for maybe 50% more at next company struggling desperately with same stuff.
There is a ton of much more ordinary software engineering work that goes into building this stuff. The people being talked about in this article could just be frontend or backend engineers who were working on a product listing page this time last year.
“Fuck you, pay me”
Not sure why this is so hard. Makes it even difficult when you have bendy co-workers.
You are basically saying "Pffft. First world problems." Which is always a flippant dismissal of someone's actual problem, and you're doing it in HN of all places. A website by YCombinator, which is more or less the tip of the spear of the first world.
AI is making even more established companies behave like early startups. shifting deadlines, building things to appease investors and fight for funding, learn how to fly the plane while building the engine, having to set your own boundaries, etc.
It’s always strange for me to hear anecdotes from “not-early-startup” folks complaining about not knowing how to set boundaries or needing learning to be spoonfed to them when they’re exposed to an early-startup-like environment.
I always wonder if they’re in the majority or I am.
https://www.makeartwithpython.com/blog/what-is-happening-wit...
tldr; too much capital necessary for wide participation, secrecy and paranoia is a key part of building a moat, and questionable training data. the high salaries just add to the pressure.