Amazon Employees Say AI Is Just Increasing Workload | Hacker News Reader

Amazon Employees Say AI Is Just Increasing Workload

gizmodo.com

123 pointsjamesgill38 comments

Not seeing this yet but our place is still trying to figure things out. We have strong KPIs on how much we use copilot so I must say woof woof to it every day or so to make sure it shows me as an active user (luckily it only shows the latest date someone has used it in each application).

For programming I find it pretty useful. For MS Office it's so far not a hell of a lot more useful than clippy. The one thing it's good at is finding old stuff in SharePoint and Outlook but that's more a sign of how terrible the search functions are in those than of how good copilot is.

A friend of a friend has too - they just ask the AI some random questions in the morning to hit their KPIs because AI doesn't make for a good accountant.

That’s far from the truth. LLMs are excellent accountants, especially for net zero balance stuff which is how most companies run.

Maybe - I think my friend's friend is more in the giving nuanced advice space of accounting rather than anything too mathematical.

I know accountants, most of the stuff they deal with is getting stuff to make sense when your dealing with an unreliable narrator dealing with push back from their employers and then attempting to synthesize a coherent narrative over a large corpus of heterogeneous internal documents, controls, spreadsheets and legal agreements. LLM’s as they exist now are yes men, they are not good at making coherent fact patterns, especially when someone pushes back on them or gives them a red herring or mis-represents something. There’s a long list of things that need to improve for them to gain traction in that industry.

I literally say woof woof hehe. I have to be careful sharing my screen because the sidebar (conversation history) is full of "morning puppy greetings!!" and stuff.

I can see the reports being generated myself (I work in that part) so I know what I'm doing looks good enough.

But it would be funny if I really had to give it busywork. I'd probably build a local agent (like with foundry or something) to talk to copilot keeping it busy. I like the paradox of that.

>We have strong KPIs on how much we use copilot so I must say woof woof to it every day or so to make sure it shows me as an active user (luckily it only shows the latest date someone has used it in each application).

Ask it to generate a cron job for this.

that would just create more evidence of him not using it regularly. not good

This is happening to me too. I am so productive now, with this super power, that my reward is to be given more work.

That's because everyone else has access to the same tools. If you don't become more productive then you will be replaced with someone who will.

They only way it'd help you is if you controlled access to AI and it was a competitive advantage for you over a fellow developer. A rising tide lifts all boats... but you are only paid for how much taller you are than the other boats.

It's important to understand that AI is capital, and it's to the owner of the capital go the spoils. Your capital is in your skills, but they are a commodity and thus you have limited leverage.

Who benefits from AI is smaller businesses who could not afford custom application development at previous development costs. It's like faster laptops and better IDEs didn't boost developer salaries.

>Who benefits from AI is smaller businesses who could not afford custom application development at previous development costs.

Of course, as AI reduces the cost to operate in niches, those small businesses who just gained the ability to build an app are also more likely than before to see a bigger player drink their milkshake.

Not to mention that small businesses will have a harder time absorbing the inevitable price hike that will come once everyone has made themselves completely dependent on AI to get any work done.

VSCode has advanced editor functionalities that not many use. But devs using that functionality aren't replacing devs not using it. Although they also save hours of manual work thanks to multi line cursors

Thing is, AI is being shoved top-down on devs because it's a goal for everyone, without checking if it's actually helping or hindering.

What I am seeing is juniors are blindly trusting AI output and say in a few years, it's going to be a disaster because nobody understands anything except seniors

  > What I am seeing is juniors are blindly trusting AI output and say in a few years, it's going to be a disaster because nobody understands anything except seniors

I see exactly the same in teams of senior and staff engs, and I promise you - they are not going to understand anything either.

Ouch. If seniors and staffs do this then the company is doomed.

25 years ago, a recently hired junior accountant would just run a cost accounting report and hand it off as if it were golden. Those of us who knew the system(s) would check the results and often provide some interpretation or even hold a report back so as not to mislead.

Plus ça change…

Well yeah, that’s the whole point. My company is still at the phase where people have access to LLMs but it’s in the honeymoon phase where expectations for work throughput are still pre LLM but everyone secretly uses them and slacks off the rest of the time.

I'm not slacking! I'm recharging.

My issue is that now have a super effective rubber duck.

The friction that I had in the past to getting started is gone. I have something to work on 30” after I sit on my desk.

And if something is not working, I have someone eager to discuss with me about it and patiently work towards a solution.

One positive consequence of AI is for people working on old, constantly updated codebases. Especially the stuff created in a data scientist development paradigm (my adhoc python script produces good results, let me clean up a bit and merge into prod codebase).

There's suddenly much more interest in refactoring, test coverage, etc. and more space for this work, both because it enables more AI work and because AI on clunky code makes it even clunkier much faster than human developers (who are not data scientists ;))

In addition AI makes it easier. Tell me which ones of the 70 fields in this monster class are not used for anything of consequence anymore, this kind of stuff .

Yea. They are half assed tools that can do _something_ but aren’t a magic bullet, yet executives put guidelines in like increase productivity by 100% because they either believe or, want to manifest into reality, that they are a magic bullet for the deadly sin of paying employees.

So now expectations are based on a false reality and everyone has to work harder.

interested in what part - i found making qcli agents to run through SOPs was real effective and turned 20min of busy work async.

there might be a misattribution though. over time, id say the top down initiatives and quality at tracking them has gone up maybe 100% used to be one big project for getting rid of oracle that went for 3-7 years, but nowadays theres maybe 5 per month that disrupts the whole company at once.

There's a desire to make that LLM/agent based, but the agent still doesnt cover all the communication overhead to actually communicate the change across teams for deployments, nor is it so seamless that you dont have to take time out to understand whats happening and schedule the work, even if it is sometimes just approving an automated code change

> maximizing profit for shareholders

This is the only phrase where I see where the concern for others is so deeply and genuinely expressed in america.

This year is either going to show that LLMs are really going to be super-transformative, or, the investment thesis is a basket-case.

Strap in.

> The AI tools ultimately helped the company in its quest for more output, but didn’t help the employee who is looking to ease her work burden

That seems like it’s doing exactly what it is supposed to. Did people think that AI would give them a four hour work day? Have you not lived under capitalism for years now?

Of course it does, the same way that industrial revolution increased it. Those that tell you about automation by itself is leading to better quality of life, better working conditions etc. either are ignorant, or simply lie to you. All the good stuff we have was obtained through political means. Yes, technology was something that enabled particular outcome, but only if harnessed by political means.

The fact we can today enjoy 40 hours workweek is not a necessary consequence of steam machine. It is a consequence of people dying while fighting with police and capital henchmen in Chicago, and it other places.

But keep believing your overlords and their servants.

> keep believing your overlords and their servants.

HN is full of servants hoping to be overlords one day. For all the intelligence on this site it's amazing how many lapdogs there are here

Money makes people pathetic

Let the execs eat cake. Stop correcting the AI's mistakes.

I think developers will be paid to use AI to develop and part of that pay is to correct AI mistakes. When the AI doesn't make any mistakes is the time development takes another different turn.

This is a big when.

Duh, it’s Amazon. They love piling more work on. Why anyone still works there is beyond me unless they have no other choice. The word was out a decade ago.

Only 2 reasons one would stick around: Money and/or visa constraints

Come on. Machines that save physical labor can shorten the workday, but don't. They do lessen the calories you burn. Email, laptops and phones, all the cogntive and information inventions increase the workload. Either expectations rise (hello 3am Slack message) or employees are trimmed.

Prime age labor force participation is as high as ever

Reply on news.ycombinator.com