As someone who has spent a vast portion of life believing technology would make life better, I've come to the realisation that this idea is a fallacy. Technology amplifies power and until we collectively redefine and enforce a value system that benefits us all, the advancements in technology simply serve as a means of subjugation
I think there is a bit of wider social norms piece missing as well on AI use in knowledge work context.
Someone forwarded an enormous amount of text over teams the other day at work. From someone (bless her) that always means well but usually averages about one spelling mistake per word and rarely goes over 20 words per message. Clearly copy paste chatgpt.
For say hn gang that thinks in terms of context shifts, information load and things on THAT wave length the problem with that situation is obvious but I realised then that is not at all obvious to the average public. She genuinely seemed to think she's helping me by spending 15 seconds typing in a prompt and having me spend the next 30 minutes untangling the AI slop.
There is zero understanding or consensus of acceptable practices around that sort of thing baked into societal norms right now.
You can use an LLM to fix spelling and grammar errors. You don't need to generate slop. (Cloud providers sell LLMs as "robot information workers" when they're actually "calculators for text".)
> She genuinely seemed to think she's helping me by spending 15 seconds typing in a prompt and having me spend the next 30 minutes untangling the AI slop.
What you are seeing is a seed of the future of communication.
Communication is one of the hardest things people do. The goal it to pass idea from the sender to recipient in a manner that is least lossy. Look at how many things need to be aligned for that to even barely succeed. You have to speak the same language, dialect, have similar enough personal vocabulary, have sufficiently aligned mindsets in the domain you are communicating about, have the same current context and ability to convey context update, then the sender must serialize their ideas into actual words with correct enough spelling and in correct order to get the idea into recipients mind. All that while knowing only very little about recipients mind and having to predict what effect the words will have on it, assuming they don't misread the text.
In the future barely anyone will produce raw text themselves. At least not in professional contexts. World will be way more mixed. People will come from very different cultures and use very different languages. Most people you will encounter in professional setting will not be sufficiently aligned with you to communicate anything beyond the simplest ideas. And neither you nor them are going to be willing to align with others.
You know what will align with you? Your AI. So any message from a human will go through your AI and any message crafted by you for a human will go through your AI as well. And when it's received, through theirs. Messages will not be written. They will be constructed in a dialog with senders AI. And they won't be read. They will be interrogated in a dialog between recipient and their AI.
The future is going to be way more diverse. People will use their own communication styles they were taught when they grew up. But the bulk of out-of-family communication will be done through AI. And the AI language will be verbose. Not really fit for routine human consumption. Because words are cheap for AI and not losing details is a communication priority. It's starting as a corporatese English. But I think it will evolve rapidly to increase signal to noise ratio (while still being impractically voluminous for humans).
They issue now is just that you are trying to read rudimentary machine code of future human communication directly.
> helping me by spending 15 seconds typing in a prompt and having me spend the next 30 minutes untangling the AI slop.
Only 30 minutes? You have it good! ^_^
This person is creating more work for an FTE who now has both a) the original job, and b) the additional load of purging corruption from the inputs for (a). This is happening at scale.
Your tolerance for this depends on how close to capacity you are for (a). It's a tale as old as corporate time, well-documented by Office Space and Dilbert.
Work is Work. Pantomime is Pantomime , whether it's with "frontier" or low-tier LoLMs.
> There is zero understanding or consensus of acceptable practices around that sort of thing baked into societal norms right now.
What is acceptable right now is to believe that corrupting the inputs to the work of serious FTEs is somehow beneficial. You are expected to sing the revised words of the corporate anthem with your customary passion and obedience. Layoffs will continue until the morale of data centres has improved.
Not going to lie, I have no pity for the tech employees of a company that has spent most of its existence making the world a worse place. They are finally getting a taste of the medicine Facebook has been giving to everyone in the last 2 decades.
I hear you, but many in the tech industry will likely use this moment as a bellwether of what else they can do to extract every last drop of value from their employees. This is going to get really fucking ugly really quickly if, say, Microsoft, were to package this capability up as part of their fleet management suite to sell to companies who all want their own models like this.
> it is cutting jobs to offset its A.I. spending, saying last month that it would slash 10 percent of its work force.
> Meta also introduced internal dashboards to track employees’ consumption of “tokens,” a unit of A.I. use that is roughly equivalent to four characters of text, four people said. Some said the dashboards were a pressure tactic to encourage competition with colleagues. That led some employees to make so many A.I. agents that others had to introduce agents to find agents, and agents to rate agents, two people said.
Maybe the first to be laid off should be the ones that thought it made sense to track token consumption. Goodhart's Law doesn't even apply in this scenario because that's a dumb metric whether or not you're using it to evaluate employees.
Every big tech company's embrace of AI is making all of their employees miserable.
Whereas if you're half-competent and at a startup, the AI is an incredible opportunity to try to leap ahead while the prices are subsidized (by the big tech behemoths fighting wth each other)
The reason is a complete inversion of Ownership and Agency.
For a decade of ZIRP, big tech convinced its employees that they're "changing the world", and what we did mattered. Sure the exhorbitant salaries and constantly rising stock value didn't hurt, but honestly other than the FIRE cultists, for most of us the difference between 200k/year and 800k/year didn't feel much day to day (other than the ability to buy a house or something, and feel safe with a retirement nest egg). No, most people were missionaries not mercanaries.
2021 was the first crack. The comps went crazy, half the industry turned over, and the ones who didn't felt a bitter sting where it became blatantly clear that all the new arrivals were just in it for the $$$, and the companies were willing to pay for the backfills but not to reward the loyalty of the missionaries.
Then came the yearly layoffs, chipping away further, and reminding every employee that they're at the mercy of a spreadsheet and the whims of people 3 levels above them in the org chart, in spite of the economic reality of their product, or their personal productivity.
And now we're here, and it's clear that all of the above is still relevant. The old-timers that hung around see that their personal output doesn't matter, their product's PnL doesn't matter. All that matters is 1) the company's AI strategy (and if they're not part of it, they're secondary), and 2) tokenmaxing.
How can anyone find joy in this environment unless they're purely in it for the comp?
I couldn't. I left my big tech job in December after 15 years, and have not been this happy at work since pre-COVID.
It occurred to me recently that AI's degradation of the human factor via way of increased pressure on the remaining ranks of humans might actually be far more damaging than the AI's output itself.
from a different perspective, there are way more people who are truly miserable these days comparing to these who earn probably more than half a million per year on average. we must live in parallel universe.
On top of token tracking, they're also scoring employees on how much they teach Ai to their colleagues. As bad as the token dashboard sounds, employees being forced to try to mine each other for credit sounds worse.
Modern elites forgot that treating workers nicely was the compromise we as a society settled on because the alternative is pitchforks and torched homes.
In 2019 I suggested[0] you might reach AGI if you train on computer usage - mouse movement, keypresses, what's on the screen etc. - and it sounds like Meta are kind of trying some form of it.
78 comments
[ 7.2 ms ] story [ 61.4 ms ] threadSomeone forwarded an enormous amount of text over teams the other day at work. From someone (bless her) that always means well but usually averages about one spelling mistake per word and rarely goes over 20 words per message. Clearly copy paste chatgpt.
For say hn gang that thinks in terms of context shifts, information load and things on THAT wave length the problem with that situation is obvious but I realised then that is not at all obvious to the average public. She genuinely seemed to think she's helping me by spending 15 seconds typing in a prompt and having me spend the next 30 minutes untangling the AI slop.
There is zero understanding or consensus of acceptable practices around that sort of thing baked into societal norms right now.
What you are seeing is a seed of the future of communication.
Communication is one of the hardest things people do. The goal it to pass idea from the sender to recipient in a manner that is least lossy. Look at how many things need to be aligned for that to even barely succeed. You have to speak the same language, dialect, have similar enough personal vocabulary, have sufficiently aligned mindsets in the domain you are communicating about, have the same current context and ability to convey context update, then the sender must serialize their ideas into actual words with correct enough spelling and in correct order to get the idea into recipients mind. All that while knowing only very little about recipients mind and having to predict what effect the words will have on it, assuming they don't misread the text.
In the future barely anyone will produce raw text themselves. At least not in professional contexts. World will be way more mixed. People will come from very different cultures and use very different languages. Most people you will encounter in professional setting will not be sufficiently aligned with you to communicate anything beyond the simplest ideas. And neither you nor them are going to be willing to align with others.
You know what will align with you? Your AI. So any message from a human will go through your AI and any message crafted by you for a human will go through your AI as well. And when it's received, through theirs. Messages will not be written. They will be constructed in a dialog with senders AI. And they won't be read. They will be interrogated in a dialog between recipient and their AI.
The future is going to be way more diverse. People will use their own communication styles they were taught when they grew up. But the bulk of out-of-family communication will be done through AI. And the AI language will be verbose. Not really fit for routine human consumption. Because words are cheap for AI and not losing details is a communication priority. It's starting as a corporatese English. But I think it will evolve rapidly to increase signal to noise ratio (while still being impractically voluminous for humans).
They issue now is just that you are trying to read rudimentary machine code of future human communication directly.
Only 30 minutes? You have it good! ^_^
This person is creating more work for an FTE who now has both a) the original job, and b) the additional load of purging corruption from the inputs for (a). This is happening at scale.
Your tolerance for this depends on how close to capacity you are for (a). It's a tale as old as corporate time, well-documented by Office Space and Dilbert.
Work is Work. Pantomime is Pantomime , whether it's with "frontier" or low-tier LoLMs.
> There is zero understanding or consensus of acceptable practices around that sort of thing baked into societal norms right now.
What is acceptable right now is to believe that corrupting the inputs to the work of serious FTEs is somehow beneficial. You are expected to sing the revised words of the corporate anthem with your customary passion and obedience. Layoffs will continue until the morale of data centres has improved.
> Meta also introduced internal dashboards to track employees’ consumption of “tokens,” a unit of A.I. use that is roughly equivalent to four characters of text, four people said. Some said the dashboards were a pressure tactic to encourage competition with colleagues. That led some employees to make so many A.I. agents that others had to introduce agents to find agents, and agents to rate agents, two people said.
Maybe the first to be laid off should be the ones that thought it made sense to track token consumption. Goodhart's Law doesn't even apply in this scenario because that's a dumb metric whether or not you're using it to evaluate employees.
Whereas if you're half-competent and at a startup, the AI is an incredible opportunity to try to leap ahead while the prices are subsidized (by the big tech behemoths fighting wth each other)
The reason is a complete inversion of Ownership and Agency.
For a decade of ZIRP, big tech convinced its employees that they're "changing the world", and what we did mattered. Sure the exhorbitant salaries and constantly rising stock value didn't hurt, but honestly other than the FIRE cultists, for most of us the difference between 200k/year and 800k/year didn't feel much day to day (other than the ability to buy a house or something, and feel safe with a retirement nest egg). No, most people were missionaries not mercanaries.
2021 was the first crack. The comps went crazy, half the industry turned over, and the ones who didn't felt a bitter sting where it became blatantly clear that all the new arrivals were just in it for the $$$, and the companies were willing to pay for the backfills but not to reward the loyalty of the missionaries.
Then came the yearly layoffs, chipping away further, and reminding every employee that they're at the mercy of a spreadsheet and the whims of people 3 levels above them in the org chart, in spite of the economic reality of their product, or their personal productivity.
And now we're here, and it's clear that all of the above is still relevant. The old-timers that hung around see that their personal output doesn't matter, their product's PnL doesn't matter. All that matters is 1) the company's AI strategy (and if they're not part of it, they're secondary), and 2) tokenmaxing.
How can anyone find joy in this environment unless they're purely in it for the comp?
I couldn't. I left my big tech job in December after 15 years, and have not been this happy at work since pre-COVID.
Ford style assembly lines made the work of the factory workers more miserable. Partially automated cashier did the same thing.
I don't think there is any point in trying to resist automation, as the efficiency benefits are too important.
https://www.nbcdfw.com/news/nbc-5-responds/meta-users-contin...
I say this as someone self employed that burned almost $1000 on tokens last month. And had. A lot of fun doing it.
0. https://svilentodorov.xyz/blog/human-imitating-task/