Tech companies are scared to death of unionized workers. But they relay their development to AI companies that will increase the price of their services all at once, as much a they want.
We've reached the point in capitalism where companies make money off of speculation rather than actually making goods or providing services. So these companies know exactly what they are doing. They don't care if their products are shitty, they just need the hype to get that sweet sweet investment money (your retirement funds).
When making money becomes the prime motivator of a company, that company will eagerly sacrifice any usefulness it had in exchange for greater profit margins. Like a pig realizing it's made of food, so why eat something else?
Sure, but there's a reason olympic runners aren't jumping at the chance to replace their legs with prosthetics for a chance to improve their times. Si non confectus, non reficiat.
An AI will not replace a programmer completely today. But it does make programmers more productive. If a team of 10 programmers is suddenly 20% more productive, the company will need to think if it's happy about the increased output or can cope with 8 programmers with the same output instead.
It's not impossible to saw off a limb friend. And also looking into it IT IS RECOMMENDED AGAINST DOING THIS. So even if I take you literally not literally and concede that point... Its just as stupid an idea. Increasing your power is more important than losing weight ESPECIALLY at the cost of muscle mass. So your comment is completely fucktangular.
I have a very simple gpt at work that's just supposed to look at a file in its knowledge base, go throughthe questions, ask me for answers, then uses the answers to spit out an email template. Other day chatgpt told me (didn't ask) it will use "work mode" going forward to think more on problems. I try to use my little gpt today and the damn think was adlibbing shit like crazy. Had to waste time redoing the instructs on it to make it act like it used to. AI is getting sketchy af
This is one of the biggest reasons why I've always refused to use it. If the company can change the model at any time without me knowing or understanding how
Well I mean, you can refuse to use a bad harness for that reason I suppose, but good harnesses allow LLM selection. The harness itself can change if you update it, but you can control that too. And if you're referring to business solutions using LLMs you just integrate directly and configure your hosting environment to never auto upgrade the model.
This isn't a good reason for refusal most of the time.
I mean, it's not the only reason, I also just don't have a lot of good reasons TO use a LLM. The things that I do outside of my work that it COULD help with are the things that I enjoy doing, so using a LLM to accomplish the tasks would rob me of the enjoyable activity.
It's just part of the host of reasons to not use a LLM
Sure there are good reasons. I took issue with the reason you gave, which was not one of the many valid ones as I outlined in excruciating, tedious detail lol.
Just getting a new android phone set up takes me 5+ hours. Just loaded with crap (ai, child mode, two separate battery savers, undeletable FB, Google search and news) and impossible permissions (ghosted alarm permissions for podcatcher, etc.) The idea that something I depend in for work could change that wildly is horrifying. Luckily I'm in manufacturing so our machines run NT
I remember when the Brawndo stock went to 0 because Not Sure started putting toilet water on all the crops and the auto-layoff AI fired all the employees.
That's going to be the ONE way we get these dipshits to stop shoving AI down our throats is if the parasite class starts feeling like their income is threatened.
Yes, a company with a decent C-suite does way better than a company there the C-suite are only there to collect a paycheck. You just don't see decent C-suites very often these days
Well, if most real CEOs are themselves parodies of actually good CEOs, then it follows that AI CEOs being parodies of actually good CEOs wouldn't be much different.
Another sign of the disconnect between management and engineering is that as a software developer of 20+ years, there has never been a time where my to-do list hasn't basically been endless. AI could have been a good tool to get more stuff done with the same number of people, but instead they've created a situation where productivity will go down (and AI costs are rising)
Every tech company is run by some jackass business idiot MBA, and none of them have a goddamn clue how to run a tech company properly, just to run one profitably.
I worked a security job with nearly two decades of experience, and got sick. Dues paid, they said, and sat me up with a light duty job watching cameras and the like. Two years of sitting in a tiny box all alone for 12 hours a day, sometimes 6 or 7 days a week, then a new recruitment incentive would result in me getting a $10k raise. They drummed me out, and within the next few years they got countless lawsuits due to inexperienced and undertrained staff.
Also got 20 years experience mainly in javascript and php. Cannot find a job in the UK. Seeing 200- 500 applications to jobs when i ask in the rare times i interview. After 12 months looking as a self employed contractor i'm done. I'm out the game. Good luck. Tax the rich.
Making management it's own class of employee with no practical knowledge of what they are managing is a terrible idea. Get an accountant and a HR person and let your owner/qualified personnel actually make business decisions. The C suite is all leaches.
I’m divided on the usefulness; there’s so much evidence of people blindly accepting sloppy code that will quickly become technical debt that I’m not sure about the gains in those areas where it’s actually useful.
There's a lot of useful code written by AI. I run an open source project and I get a bunch of code changes ("PRs") so obviously AI written and it's usually good. Problem is there's TOO MANY for me to manage and I don't want to hand control over to AI
Remember the `xz` attack? The lone maintainer of an infrastructure project gets buried in PRs, then a "helpful" dev appears, and after some time, sneaks a backdoor into the code? And it was only discovered by a combination of dumb luck and weaponized autism? Yeah, LLMs are basically facilitating that, at scale.
Keep your project focused, eschew bloat and feature-creep, and trust no one.
It's only useful when it's overseen by someone who knows exactly what they're doing and how to set up and manage the correct documentation for it to follow. I'm not defending AI usage at all but just pointing out that it can be used effectively. Ethically? Less likely, unless you're using local only models, and only for code output.
This is what I was thinking of as well, but you're right I should have probably specified this too. It needs competent / knowledgeable people at all points of use and review.
Yeah, it's great when you need a quick function refactor to better handle things like transient network drops and retry loops. But the key is knowing both what to ask for, and how to check the output for correctness. It's also great at quickly processing a stack trace
Reminds me of the java explosion. Tons of code monkeys were hired,,and they cranked out shit code. Then companies spent the next decade hiring real architects to refactor and optimze it all. AI is just the next wave of code monkeys shitting every where, and companies are starting to realize they need to bring back the engineers to fix it all again.
Worse. In previous advancements you'd have 90% trash programmers, but 10% who actually learned and got good, and were usually retained long after the initial development interest faded to. Now you get 100% trash, and actively LOSE competent programmers as they're replaced with shitty AI that seems to be getting worse not better, as time goes on.
Definitely not 100% trash, since many competent programmers use AI to help writing code. They're not being replaced, they're being helped. What makes you think that it's "shitty AI" that's getting worse? Have you done any vibecoding yourself to see what it can do?
A smart person would create a startup that hires true experience, AKA actual top notch senior devs, and graciously offers to unfuck their AI bullshit once they hit a bump. That way there won't be a huge pool of people they can just fire and rehire.
At least with hiring code monkeys you'll likely have a few of them actually getting competent and thus adding to the workforce of engineers who can fix stuff. The current trend doesn't even raise any new talents and if anything has resulted in the deskilling of folks who used to be competent.
I've read actual accounts of devs switching to "vibe coding" being unable to solve any issue that arises since they haven't actually been writing the code
Oh no. You're expected to get more stuff done with less people. I'm in this picture. People have left and they have not been replaced. Juniors and students have not been hired. Some products are worked on by a single person and are expected to do more than ever before in a short time frame. We're not getting new people because they cost too much but small teams burning through 50.000€+ of tokens per month is fine.
Think of it like tickets at a fairground. You buy tokens with real money, and that buys you time with the AI - just like you give tickets to ride a ride. The problem is, much like tickets at a fair, the price is completely arbitrary and can change at any time, and what time an amount of tokens can buy can change at any time too. It also has the benefit of being a currency you can ration, for better or worse, rather than a monetary budget.
This is one of the biggest concerns no one at my company is listening to. Ai companies are not making money yet. When they do it will be through massive price increase. And if you fired people, you are in a real tough spot.
You know how predatory phone games will lock stuff like loot boxes behind a fake currency like "gems" so the actual cost isnt as obvious? Its like that but for shareholders. Saying "our devs used 2300 AI tokens to produce this tool that will generate $50,000 in income." Sounds a hell of a lot better than "our devs spent $230,000 to produce this tool that we think will make us $50,000."
When an AI reads or writes a text, the AI breaks the text down into small chunks of usually single words or even parts of a word which it recognizes as the smallest unit of text that holds information. These chunks are called tokens. Depeding on the AI model, the number of tokens a text is composed of may vary. Some AI models may interpret e.g. the word "fundamentally" into two tokens ("fundamental" and "ly"), a different AI may interpret it as only one. Based on the AI model, a token will
incur different costs, with more advanced models usually having higher costs per token as older or less advanced ones. There's also a distinction between input and output tokens, as generating output tokens is far more complex than input tokens.
Think of a token as four letters of the alphabet. Your message is broken into tokens. The response is broken into tokens. You're billed for both. Any intermediate steps, if the AI is calling it's own AI (this is an agentic loop bc AI agent is looping with it's own logic) those will also be billable tokens
Which usually burns a lot of useless tokens and ends up costing so much more. Luckily, most show their thought process so you can see it spiraling down a path you don't want and can stop it before it does something really stupid.
Tokens are, in essence, word combinations generated by LLMs, and is currently how they're priced in usage. So a few hundred thousand tokens is a few A4 papers worth of generated output. A million a few more A4 papers, etc etc. Taking some shortcuts here but if someone more tech savvy would fill in I am sure my friend up here and me will appreciate it.
This is incorrect math. A page single spaced is about 500 words. Average word 5 letters. That's 2500 letters per page. A token is approximately 4 letters. 625 tokens per page. A million tokens is approximately 1,600 pages. And the cheapest Claude plan is a dollar for a million tokens.
Thank you for the correction, I was off by a mile but the general concept stands. However I am curious when, if ever, we get to see more transparent pricing because it depends an awful lot on what you're asking the LLM to generate. I do not recall token generation being letter based, but more like prefix/suffix/postfix based per-word? I'll have to read up on it and I'll try to share my own correction on my understanding of token generation.
Not too sure, as the next step would be to replace your limb with an artificial one (Flex-Foot Cheetahs), which according to popular myth, would make you run faster.
It’s not my fault you and the downvoters are ignorant asses with no critical thinking skills. Comic does not contain a “replacement.” Just a removal of the limb. So it just inaccurate. Factually and unarguably. Just a stupid comparison to begin with. Then it makes the premise that an amputee is less than or slower because of that disability. Using a disability in a factual inaccurate comparison to negate an amputee’s abilities for a comic is as textbook of a definition of ableism as there is.
That is not a disability, that is stupidity disguissed as progress. The person cutting off the leg is not going to be faster, now is just mutilated. The thing here is that AI is not a good replacement, firing people to put AI in charge is stupid, you are just making everything less reliable and more difficult for everyone.
So cutting a limb for profit is not the same as a dissability, and AI is not considered a replacement, just a downgrade.
No. Your point holds no sway. Yes it’s a disability. Or a limited ability. Or whatever the best euphemism is nowadays. They’re an amputee. How they ended up that way would only be a symptom of mental illness, one that has actually happened, so the comic also just makes light of that. Yes we get AI sucks, and this comic is trying to humanize it. It did it in a shitty illogical way.
And fuck that noise “now they’re just mutilated.” No. That’s the problem. Amputees are not “just” mutilated. It’s limited scope and dismissive to presume. Again the comic is not saying “remove” the dev it’s “replacing them”. The comic is incomplete and the leg has to be replaced with a highly advanced prosthetic to represent AI, so that even if they can still go fast, they might end hip running directly into a wall. The comic sucks.
demosteness
Tech companies are scared to death of unionized workers. But they relay their development to AI companies that will increase the price of their services all at once, as much a they want.
SpanWolf
Look how much easier it is to pull up my bootstraps.
TheJuiceLoosener
txinjing9000
We've reached the point in capitalism where companies make money off of speculation rather than actually making goods or providing services. So these companies know exactly what they are doing. They don't care if their products are shitty, they just need the hype to get that sweet sweet investment money (your retirement funds).
etherbunny41
When making money becomes the prime motivator of a company, that company will eagerly sacrifice any usefulness it had in exchange for greater profit margins. Like a pig realizing it's made of food, so why eat something else?
Urgalicity
This is just how management thinks about everything.
OperatorWay
kInADress
At this point just fucking go ahead, nuke your company, see if I care.
landbaronness42
Ford replaced veteran engineers with AI. Recalls skyrocketed.
https://www.autoblog.com/news/ford-learned-the-hard-way-that-ai-cant-replace-veteran-engineers
demosteness
Exactly
Bizarkly
Should have shown him cutting off his head: "I can think more clearly now!" But this is good enough to *make the point*: well-done, @OP!
Kingstad
you could argue with this analogy that if you replace your meat leg with a robot leg then it could well be an upgrade
etherbunny41
Sure, but there's a reason olympic runners aren't jumping at the chance to replace their legs with prosthetics for a chance to improve their times. Si non confectus, non reficiat.
johnathanrossshoulddie
"if you change the analogy it would mean a different thing" brilliant take bud
InexplicablyAvailableUsername
more like replace it with a leg made of cake
malachilenomade
IMakeLotsOfReferencesAndRemakes
One that repeatedly kicks you in the balls and nothing else.
WheelieNelson420
This is exactly how it is when executed badly.
An AI will not replace a programmer completely today. But it does make programmers more productive. If a team of 10 programmers is suddenly 20% more productive, the company will need to think if it's happy about the increased output or can cope with 8 programmers with the same output instead.
hendawg2020
https://media0.giphy.com/media/v1.Y2lkPWE1NzM3M2U1dW85cmN0M2kyMjBweThhbzkzb3ZweW90bTd6MmtpM2F3eWxkcWw5NiZlcD12MV9naWZzX3NlYXJjaCZjdD1n/2QdnjbwtQqRMY/200w.webp
somebackup
this is literally what climbers do, they sacrifice muscle mass to be lighter and climb faster.
etherbunny41
hello gross misinterpretation of a post!
somebackup
This is the smartest reply I've gotten, kudos to you.
etherbunny41
IMakeLotsOfReferencesAndRemakes
Show me one climber who has cut off their leg for the sake of climbing. I will wait.
somebackup
hello gross misinterpretation of a comment!
IMakeLotsOfReferencesAndRemakes
No it isn't. You said "This is >literally< what climbers do" in response to a cut off leg. There is no comment you responded to only the post.
somebackup
https://www.merriam-webster.com/dictionary/literally
IMakeLotsOfReferencesAndRemakes
It's not impossible to saw off a limb friend. And also looking into it IT IS RECOMMENDED AGAINST DOING THIS. So even if I take you literally not literally and concede that point... Its just as stupid an idea. Increasing your power is more important than losing weight ESPECIALLY at the cost of muscle mass. So your comment is completely fucktangular.
sadurdaynight
I have a very simple gpt at work that's just supposed to look at a file in its knowledge base, go throughthe questions, ask me for answers, then uses the answers to spit out an email template. Other day chatgpt told me (didn't ask) it will use "work mode" going forward to think more on problems. I try to use my little gpt today and the damn think was adlibbing shit like crazy. Had to waste time redoing the instructs on it to make it act like it used to. AI is getting sketchy af
TheZerax
This is one of the biggest reasons why I've always refused to use it. If the company can change the model at any time without me knowing or understanding how
geekykeycap
Well I mean, you can refuse to use a bad harness for that reason I suppose, but good harnesses allow LLM selection. The harness itself can change if you update it, but you can control that too. And if you're referring to business solutions using LLMs you just integrate directly and configure your hosting environment to never auto upgrade the model.
This isn't a good reason for refusal most of the time.
TheZerax
I mean, it's not the only reason, I also just don't have a lot of good reasons TO use a LLM. The things that I do outside of my work that it COULD help with are the things that I enjoy doing, so using a LLM to accomplish the tasks would rob me of the enjoyable activity.
It's just part of the host of reasons to not use a LLM
geekykeycap
Sure there are good reasons. I took issue with the reason you gave, which was not one of the many valid ones as I outlined in excruciating, tedious detail lol.
TheZerax
how can I trust the output it gives, ever? It could receive new instructions or training or data and I'd never know until it's just WRONG.
astrangehop
Just getting a new android phone set up takes me 5+ hours. Just loaded with crap (ai, child mode, two separate battery savers, undeletable FB, Google search and news) and impossible permissions (ghosted alarm permissions for podcatcher, etc.) The idea that something I depend in for work could change that wildly is horrifying. Luckily I'm in manufacturing so our machines run NT
LicensedAdHominem
Some CEO fired devs to replace them with AI so they made an AI-based CEO replacement and it's hilarious https://github.com/SenteLabsAI/OpenExecutive
AtaSeaParks
I remember when the Brawndo stock went to 0 because Not Sure started putting toilet water on all the crops and the auto-layoff AI fired all the employees.
HummingBirrb
https://media1.giphy.com/media/v1.Y2lkPWE1NzM3M2U1ZzY3MWo2enF2a2FqNGx1dzRhZWN4NXNvcHgxbHA0dDk4YTJodDRnOSZlcD12MV9naWZzX3NlYXJjaCZjdD1n/CoDp6NnSmItoY/200w.webp
Beefkins
That's going to be the ONE way we get these dipshits to stop shoving AI down our throats is if the parasite class starts feeling like their income is threatened.
Circosys
I mean if any job should be replaced by an advanced algorithm....
geekykeycap
Advanced? Eh. I bet you can run it on a calculator ;)
Penelelop
Have been saying this since the start
Cilvaa
Is it just a parody of a CEO, or does it do actual work?
AtaSeaParks
If a placebo works...
reichstein
Would there be any difference?
QuartzPoker
Yes, a company with a decent C-suite does way better than a company there the C-suite are only there to collect a paycheck. You just don't see decent C-suites very often these days
etherbunny41
Well, if most real CEOs are themselves parodies of actually good CEOs, then it follows that AI CEOs being parodies of actually good CEOs wouldn't be much different.
Skuggen
Another sign of the disconnect between management and engineering is that as a software developer of 20+ years, there has never been a time where my to-do list hasn't basically been endless. AI could have been a good tool to get more stuff done with the same number of people, but instead they've created a situation where productivity will go down (and AI costs are rising)
SavageDrums
Every tech company is run by some jackass business idiot MBA, and none of them have a goddamn clue how to run a tech company properly, just to run one profitably.
Draygen
I worked a security job with nearly two decades of experience, and got sick. Dues paid, they said, and sat me up with a light duty job watching cameras and the like. Two years of sitting in a tiny box all alone for 12 hours a day, sometimes 6 or 7 days a week, then a new recruitment incentive would result in me getting a $10k raise. They drummed me out, and within the next few years they got countless lawsuits due to inexperienced and undertrained staff.
somebackup
I don't even want to look at what my backlog is at. Few hundred. Once a year I go and cross things off as never happening.
JohnSmithterms
Also got 20 years experience mainly in javascript and php. Cannot find a job in the UK. Seeing 200- 500 applications to jobs when i ask in the rare times i interview. After 12 months looking as a self employed contractor i'm done. I'm out the game. Good luck. Tax the rich.
pgdave
I've had pretty good results using AI to do some of the easier tasks and scripting tools to make my life easier, rather than directly coding.
Magus25
Making management it's own class of employee with no practical knowledge of what they are managing is a terrible idea. Get an accountant and a HR person and let your owner/qualified personnel actually make business decisions. The C suite is all leaches.
nihiltres
I’m divided on the usefulness; there’s so much evidence of people blindly accepting sloppy code that will quickly become technical debt that I’m not sure about the gains in those areas where it’s actually useful.
rhalm
That...that's literally been the case since the dawn of coding.
shitheadtookmyname
There's a lot of useful code written by AI. I run an open source project and I get a bunch of code changes ("PRs") so obviously AI written and it's usually good. Problem is there's TOO MANY for me to manage and I don't want to hand control over to AI
0xDEC0DE
Remember the `xz` attack? The lone maintainer of an infrastructure project gets buried in PRs, then a "helpful" dev appears, and after some time, sneaks a backdoor into the code? And it was only discovered by a combination of dumb luck and weaponized autism? Yeah, LLMs are basically facilitating that, at scale.
Keep your project focused, eschew bloat and feature-creep, and trust no one.
satanslittleposter
It's only useful when it's overseen by someone who knows exactly what they're doing and how to set up and manage the correct documentation for it to follow. I'm not defending AI usage at all but just pointing out that it can be used effectively. Ethically? Less likely, unless you're using local only models, and only for code output.
nihiltres
I’d take it a step further and say that even a competent person needs to *actually do* the necessary review.
satanslittleposter
This is what I was thinking of as well, but you're right I should have probably specified this too. It needs competent / knowledgeable people at all points of use and review.
Banjo143
Yeah, it's great when you need a quick function refactor to better handle things like transient network drops and retry loops. But the key is knowing both what to ask for, and how to check the output for correctness. It's also great at quickly processing a stack trace
sadurdaynight
Reminds me of the java explosion. Tons of code monkeys were hired,,and they cranked out shit code. Then companies spent the next decade hiring real architects to refactor and optimze it all. AI is just the next wave of code monkeys shitting every where, and companies are starting to realize they need to bring back the engineers to fix it all again.
rhalm
So this whole scenario with AI is the same as previous computer development advancements?
DrBlackJack
Worse. In previous advancements you'd have 90% trash programmers, but 10% who actually learned and got good, and were usually retained long after the initial development interest faded to. Now you get 100% trash, and actively LOSE competent programmers as they're replaced with shitty AI that seems to be getting worse not better, as time goes on.
rhalm
Definitely not 100% trash, since many competent programmers use AI to help writing code. They're not being replaced, they're being helped. What makes you think that it's "shitty AI" that's getting worse? Have you done any vibecoding yourself to see what it can do?
DrBlackJack
Oh, sorry. Didn't realise I was talking with a troll. I took your original comment seriously. My mistake.
emu314159127001
A smart person would create a startup that hires true experience, AKA actual top notch senior devs, and graciously offers to unfuck their AI bullshit once they hit a bump. That way there won't be a huge pool of people they can just fire and rehire.
Colopty
At least with hiring code monkeys you'll likely have a few of them actually getting competent and thus adding to the workforce of engineers who can fix stuff. The current trend doesn't even raise any new talents and if anything has resulted in the deskilling of folks who used to be competent.
emu314159127001
I've read actual accounts of devs switching to "vibe coding" being unable to solve any issue that arises since they haven't actually been writing the code
Kodan00
Oh no. You're expected to get more stuff done with less people. I'm in this picture. People have left and they have not been replaced. Juniors and students have not been hired. Some products are worked on by a single person and are expected to do more than ever before in a short time frame. We're not getting new people because they cost too much but small teams burning through 50.000€+ of tokens per month is fine.
NotSinceTheAccidend
Tokens? I’m an outsider, what does that mean?
cyankitsune
Think of it like tickets at a fairground.
You buy tokens with real money, and that buys you time with the AI - just like you give tickets to ride a ride. The problem is, much like tickets at a fair, the price is completely arbitrary and can change at any time, and what time an amount of tokens can buy can change at any time too.
It also has the benefit of being a currency you can ration, for better or worse, rather than a monetary budget.
solsactual
This is one of the biggest concerns no one at my company is listening to. Ai companies are not making money yet. When they do it will be through massive price increase. And if you fired people, you are in a real tough spot.
ArchMagos
You know how predatory phone games will lock stuff like loot boxes behind a fake currency like "gems" so the actual cost isnt as obvious? Its like that but for shareholders. Saying "our devs used 2300 AI tokens to produce this tool that will generate $50,000 in income." Sounds a hell of a lot better than "our devs spent $230,000 to produce this tool that we think will make us $50,000."
shitheadtookmyname
Yeah 2300 AI tokens isn't 230K. Claude currently charges a few dollars for a MILLION tokens
ArchMagos
I wasn't intending to use direct comparisons. Either way, it obfuscates the true cost.
DerpMeister
When an AI reads or writes a text, the AI breaks the text down into small chunks of usually single words or even parts of a word which it recognizes as the smallest unit of text that holds information. These chunks are called tokens. Depeding on the AI model, the number of tokens a text is composed of may vary. Some AI models may interpret e.g. the word "fundamentally" into two tokens ("fundamental" and "ly"), a different AI may interpret it as only one. Based on the AI model, a token will
DerpMeister
incur different costs, with more advanced models usually having higher costs per token as older or less advanced ones. There's also a distinction between input and output tokens, as generating output tokens is far more complex than input tokens.
shitheadtookmyname
Think of a token as four letters of the alphabet. Your message is broken into tokens. The response is broken into tokens. You're billed for both. Any intermediate steps, if the AI is calling it's own AI (this is an agentic loop bc AI agent is looping with it's own logic) those will also be billable tokens
UWAGAGABLAGABLAGABA
Which usually burns a lot of useless tokens and ends up costing so much more. Luckily, most show their thought process so you can see it spiraling down a path you don't want and can stop it before it does something really stupid.
pnersvfu
Tokens are, in essence, word combinations generated by LLMs, and is currently how they're priced in usage. So a few hundred thousand tokens is a few A4 papers worth of generated output. A million a few more A4 papers, etc etc. Taking some shortcuts here but if someone more tech savvy would fill in I am sure my friend up here and me will appreciate it.
shitheadtookmyname
This is incorrect math. A page single spaced is about 500 words. Average word 5 letters. That's 2500 letters per page. A token is approximately 4 letters. 625 tokens per page. A million tokens is approximately 1,600 pages. And the cheapest Claude plan is a dollar for a million tokens.
pnersvfu
Thank you for the correction, I was off by a mile but the general concept stands. However I am curious when, if ever, we get to see more transparent pricing because it depends an awful lot on what you're asking the LLM to generate. I do not recall token generation being letter based, but more like prefix/suffix/postfix based per-word? I'll have to read up on it and I'll try to share my own correction on my understanding of token generation.
Sebastopol140
Yep, good analogy.
Raistline
Did I hear someone say Anal Orgy?!
PickleRiiiiiiiiiiiiick
Lemme ask my analtherapist.
jodu
Not too sure, as the next step would be to replace your limb with an artificial one (Flex-Foot Cheetahs), which according to popular myth, would make you run faster.
FCKYHWH
I’d say really not. You’d have to replace the amputated limb with a prosthetic. A high tech one. The comic is kinda ableist.
johnathanrossshoulddie
congrats on finding the dumbest take
FCKYHWH
It’s not my fault you and the downvoters are ignorant asses with no critical thinking skills. Comic does not contain a “replacement.” Just a removal of the limb. So it just inaccurate. Factually and unarguably. Just a stupid comparison to begin with. Then it makes the premise that an amputee is less than or slower because of that disability. Using a disability in a factual inaccurate comparison to negate an amputee’s abilities for a comic is as textbook of a definition of ableism as there is.
johnathanrossshoulddie
it isn't a contest you don't have to keep trying to outdo yourself in dumb takes
whereismymind86
Congrats on the second dumbest take
FCKYHWH
Congrats on contributing to the pit of ignorance.
Altqwenta
I'm going to bite the bait:
That is not a disability, that is stupidity disguissed as progress. The person cutting off the leg is not going to be faster, now is just mutilated. The thing here is that AI is not a good replacement, firing people to put AI in charge is stupid, you are just making everything less reliable and more difficult for everyone.
So cutting a limb for profit is not the same as a dissability, and AI is not considered a replacement, just a downgrade.
FCKYHWH
No. Your point holds no sway. Yes it’s a disability. Or a limited ability. Or whatever the best euphemism is nowadays. They’re an amputee. How they ended up that way would only be a symptom of mental illness, one that has actually happened, so the comic also just makes light of that. Yes we get AI sucks, and this comic is trying to humanize it. It did it in a shitty illogical way.
FCKYHWH
And fuck that noise “now they’re just mutilated.” No. That’s the problem. Amputees are not “just” mutilated. It’s limited scope and dismissive to presume. Again the comic is not saying “remove” the dev it’s “replacing them”. The comic is incomplete and the leg has to be replaced with a highly advanced prosthetic to represent AI, so that even if they can still go fast, they might end hip running directly into a wall. The comic sucks.