This will be a controversial opinion, but as a game developer of 14 years, AI integrations for game development are going to be standard, at least on coding.
Since we’re in PCMR, I’ll make an assumption there’s a lot of gamers here. What’s the fastest human write speed? <330wpm. What’s the fastest human read speed? Like 1000wpm. What’s the AI equivalents? We know AI can beat us on both metrics. If this was a game, what should you do? It seems like, logically, you’d focus on reading more than writing, because that’s just faster, and since the AI can beat you on the speed of writing, doesn’t it make sense to have it do that process?
The last decade or longer of IDE development has been “programmer write less.” How is AI’s writing of code much different?
Even if you were the perfect programmer, you can never beat the physical speed limits here. You can be more logically sound, make smarter decisions, but you cannot write code faster than AI.
Video games aren’t medical apps, legal apps, security apps, etc. This means lower stakes to sloppier code architecture. If a version is bad, we patch it.
Y’all… I really don’t want to admit it, but this “John Henry” line for AI is hard for me to reconcile. The one piece of solace I keep is “AI companies keep hiring human devs.” It suggests, at some level, we still need a human in the machine.
I have never felt my livelihood is under attack like this. We’re at like a +60% y/y increase of tech layoffs this year (compared to last). At the same time, there’s amazing freedom to talk to a rock and get a video game. If I was laid off tomorrow, I could use $20/mon to make a prototype I could market in like 2-3 months as a solo. That was never possible.
I’m used to reading the board state & making good decisions. This is the first time I’ve felt like the game changed dramatically around me. This sentiment is also all over my whisper network. This is more than “one person’s take.” We’re all feeling this. :(
Video game staff has always been treated like shit by the big developers.
This is another case of the big developers trying to convince their staff that using slop (and, therefore, further devaluing the people) is a good thing.
If you make a game using “AI,” you’re standing on the shoulders of companies who became giants through theft and lies. They built algorithms, the algorithms stole, and now you’re being told to use those algorithms to benefit the companies who stole to build their databases.
Yes, it sucks. You’re in a no-win situation. Things will keep getting worse, the way they’ve been getting worse for game development staff for years.
Folks here will sympathize with your personal plight - unless you use the stolen databases yourself to earn your own livelihood. Because, once you do that, you unfortunately have become part of the problem. Even if you only did it to survive.
I’m not trying to argue the sentiment, but my guess is the people that downvoted me don’t want to hear about the physical limits.
Here’s a similar version: is the human the best interface for code? I’m not convinced. Humans make mistakes when typing, we forget things. I once spent 2 days on a bug back in 2018 because an array was named “Cards” and I referenced “Card.”
Obviously, we’re the end user. The EXPERIENCE is for humans. Are we the best tool to enter code manually? I’m just not convinced.
I’m hearing stories from my engineering friends and colleagues that, for the last 6 months, 90% of their code has been AI written. I believe the engineers are smarting than me here, and that’s what they are doing.
It’s a capitalistic hellscape, bad ethics everywhere, and I’m not trying to apologize for the companies being malicious or bad actors. I’m not saying I’m morally justified. Putting that aside: what benefits do I get from writing the code myself vs describing it, if describing it generates the end result faster?
Y’all, my whole career is this field. I study philosophy and ethics undergrad. I know people don’t want to hear about it. What would you do as a rational actor in this scenario? I got 3 kids and a mortgage. :/
I agree that unions are good checks against corporate power.
How does that change the speed game we’re against? You could say, “we won’t use AI!” but your competitor will & they’ll likely go faster.
There’s a growing rift here between “chatbot for memes & my college paper” and “it coded my entire game.” Most people don’t make software, so they haven’t seen this stuff yet, or they are skeptical because ChatGPT can’t count R’s in strawberry.
I don’t know how to effectively communicate these issues right now. It feels like there are real barriers here against humans. :(
What’s the fastest human write speed? <330wpm. What’s the fastest human read speed? Like 1000wpm. What’s the AI equivalents? We know AI can beat us on both metrics.
Sometimes, quality is more important than speed.
Not that any corpo developers would understand that, since their goal is to pump out as much shit as possible as fast as possible, so they can sell more shit.
I think you’re missing my point, but if you’re gonna brand me as “corpo” without giving me a chance, then hope you have a nice day, homie. I’m not the enemy here.
I’d like to address your technical argument! I’m glad to hear you’re a study of the ethics and philosophy around this because it seems like you’re intentionally sidelining those aspects of the situation, so I want to address your technical points in a way that says, “But that’s not really the whole conversation.”
If this was a game, what should you do? It seems like, logically, you’d focus on reading more than writing, because that’s just faster, and since the AI can beat you on the speed of writing, doesn’t it make sense to have it do that process?
In general, the best thing to do in a video game is rarely, if ever, how we should approach real-world problems; those are very controlled situations that don’t cover all side-effects or consequences. I see you’re assuming a min/max strategy but that’s also not always why we play games.
This also undersells how interrelated reading and writing are. Writing makes you a better, more critical reader, and vice versa. The converse is also true. Giving up on one of the two means being worse at both reading and writing. That puts us on a bad computer literacy trajectory as a whole.
The last decade or longer of IDE development has been “programmer write less.” How is AI’s writing of code much different?
In my view it’s different because IDE development has rather been toward, “Programmer write less boilerplate”. Automating the infrastructure like getters/setters, integrating library management, version control, so that the programmer can focus on the quality of the “tangible” result of the code. With LLM tooling, no engagement is actually necessary. I’m not saying it’s better or worse, but that it’s fundamentally different to that IDE trend you suggested.
Video games aren’t medical apps, legal apps, security apps, etc. This means lower stakes to sloppier code architecture. If a version is bad, we patch it.
This is the one I come closest to agreeing with, however there’s a sticking point – even if we silo the discussion to games it’s still bad for the industry as a whole. You said yourself that you’d be able to satisfyingly monetize an LLM-assisted game in 2-3 months if you got laid off, so that signals to me that throughput is the thing you find most valuable or important about this process. What that will do is make the shovelware and malware start to look closer and closer to a quality product and make the space more dangerous to be in.
If a version is bad, you can patch it. But why would you when you can generate an entirely new game with the same level of effort? If you only need 1000 people to buy software to offset the cost, why fix it at all? Not everyone will take that stance, but many will, and that will start to make it okay for everyone. It happens even now that games release early in a buggy, incomplete state with zero hope of a fix, so I’m not sure how you can responsibly gloss over this.
Games might not be banking apps, but some games process payment information or are very invasive pieces of software and should still be scrutinized for their security practices.
In short, “why wouldn’t you?” has many answers even leaving out environmental and ethical concerns, but I’m trying to keep this response small enough to meaningfully respond to if you want.
For the ethics, I personally think these things should be nationalized / universalized, given that it runs on mass theft. That’s Meta’s argument: we stole so much, you can’t hold us legally accountable. Sure thing bud: let’s get half your profit then?
For the environmental, to me, it’s akin to driving a car. It’s very socially acceptable to do that in most places in the US. Does it contribute to worsening environment and have long term impacts? Yup, sure does. I have 1 car for our family. We used to have 2. We made a positive step, but we couldn’t lower to one. Does it make me a bad person? I don’t think so. Proportionally, I don’t think my AI usage is that impactful compared to Meta/Amazon/big tech company. Is it still bad? Yup, sure is.
That’s the reason I sidestep those, along with simplifying a complex topic I knew I would be flamed for.
Literacy matters. Seems like you’re trying to push on “what if we all did it?” And yeah, there’s some idiocracy future potential. However, I don’t think it’s addressing the issue. I mentioned John Henry because that’s partially what his story is about: man v. machine. People likely wrote faster with typewriters, and I bet people claimed literacy would go down with them. Maybe? But the reading is really essential in this process. That’s my claim.
I’ll grant there’s a degree problem with the IDE analogy, but I still think it’s a technological progression that makes rational sense.
I wouldn’t make shovelware. Many people will. I’ve said to friends I expect a Cambrian explosion. Life in all forms, good or bad. We’ll get apps that could never have been made and the worst malware to exist. It’s a Pandora’s box. The internet made us closer and brought out a lot of evil in the world. To me, internet & AI are tools, not inherently good or bad. (Philosophically, if there was no thieving of data)
I think you’re making reasonable points. I also think most people have not reconciled “it’s faster to read than to type.” It’s a strange inversion in AI workflows that comes from the mechanics at play.
Machine read speed: 1m wpm (96k per 5 secs, based on token usage of a well known model)
Machine write speed: ~6000 wpm (again, based on token usage & turnaround)
It’s not that we’re close. We’re miles apart. If we’re in a capitalistic hellscape (which, I think there’s maybe some evidence for, given on friendly convo :) ), then my competitor will beat me on speed. Why would I move a bunch of product on a hand cart vs. a train?
Thanks for the human interaction, by the way. I appreciate you engaging in earnest.
Thanks for your effort in replying too, I cherish these. Agree on the ethics and environmental points, I have nothing to add there!
I do feel that I understand where you’re coming from more now. It seemed like the focus of the argument before was just about the capacity of the tool (which is very interesting, I meant to ask for the numbers and forgot, thanks), so I’m glad we can get more nuanced.
Literacy matters! I struggled to keep this topic limited to video games, I did somewhat disregard the John Henry reference for that reason. I still don’t understand the claim here, though – are you saying that reading is more essential than writing for the purpose of literacy? It’d help me to reword that if you could.
I like the Cambrian explosion allegory because we saw something similar with the printing press and typewriter, and we certainly will here too. I’m coming from a much less realistic point-of-view than I’m sure you are as someone who works in the industry – Pandora’s Box is already open, we can’t hope to ban it or anything similar.
This will be a controversial opinion, but as a game developer of 14 years, AI integrations for game development are going to be standard, at least on coding.
Since we’re in PCMR, I’ll make an assumption there’s a lot of gamers here. What’s the fastest human write speed? <330wpm. What’s the fastest human read speed? Like 1000wpm. What’s the AI equivalents? We know AI can beat us on both metrics. If this was a game, what should you do? It seems like, logically, you’d focus on reading more than writing, because that’s just faster, and since the AI can beat you on the speed of writing, doesn’t it make sense to have it do that process?
The last decade or longer of IDE development has been “programmer write less.” How is AI’s writing of code much different?
Even if you were the perfect programmer, you can never beat the physical speed limits here. You can be more logically sound, make smarter decisions, but you cannot write code faster than AI.
Video games aren’t medical apps, legal apps, security apps, etc. This means lower stakes to sloppier code architecture. If a version is bad, we patch it.
Y’all… I really don’t want to admit it, but this “John Henry” line for AI is hard for me to reconcile. The one piece of solace I keep is “AI companies keep hiring human devs.” It suggests, at some level, we still need a human in the machine.
I have never felt my livelihood is under attack like this. We’re at like a +60% y/y increase of tech layoffs this year (compared to last). At the same time, there’s amazing freedom to talk to a rock and get a video game. If I was laid off tomorrow, I could use $20/mon to make a prototype I could market in like 2-3 months as a solo. That was never possible.
I’m used to reading the board state & making good decisions. This is the first time I’ve felt like the game changed dramatically around me. This sentiment is also all over my whisper network. This is more than “one person’s take.” We’re all feeling this. :(
Video game staff has always been treated like shit by the big developers.
This is another case of the big developers trying to convince their staff that using slop (and, therefore, further devaluing the people) is a good thing.
If you make a game using “AI,” you’re standing on the shoulders of companies who became giants through theft and lies. They built algorithms, the algorithms stole, and now you’re being told to use those algorithms to benefit the companies who stole to build their databases.
Yes, it sucks. You’re in a no-win situation. Things will keep getting worse, the way they’ve been getting worse for game development staff for years.
Folks here will sympathize with your personal plight - unless you use the stolen databases yourself to earn your own livelihood. Because, once you do that, you unfortunately have become part of the problem. Even if you only did it to survive.
I’m not trying to argue the sentiment, but my guess is the people that downvoted me don’t want to hear about the physical limits.
Here’s a similar version: is the human the best interface for code? I’m not convinced. Humans make mistakes when typing, we forget things. I once spent 2 days on a bug back in 2018 because an array was named “Cards” and I referenced “Card.”
Obviously, we’re the end user. The EXPERIENCE is for humans. Are we the best tool to enter code manually? I’m just not convinced.
I’m hearing stories from my engineering friends and colleagues that, for the last 6 months, 90% of their code has been AI written. I believe the engineers are smarting than me here, and that’s what they are doing.
It’s a capitalistic hellscape, bad ethics everywhere, and I’m not trying to apologize for the companies being malicious or bad actors. I’m not saying I’m morally justified. Putting that aside: what benefits do I get from writing the code myself vs describing it, if describing it generates the end result faster?
Y’all, my whole career is this field. I study philosophy and ethics undergrad. I know people don’t want to hear about it. What would you do as a rational actor in this scenario? I got 3 kids and a mortgage. :/
So you are not writing types as a 14 yoe developer? I am not taking your words seriously
14 yr in game design:
But yes, please write me off for some arbitrary line in the sand you drew…
What would I do? Work to form a union, probably.
No, it’s not fair. But that’s the corporate hellscape we’re in.
I agree that unions are good checks against corporate power.
How does that change the speed game we’re against? You could say, “we won’t use AI!” but your competitor will & they’ll likely go faster.
There’s a growing rift here between “chatbot for memes & my college paper” and “it coded my entire game.” Most people don’t make software, so they haven’t seen this stuff yet, or they are skeptical because ChatGPT can’t count R’s in strawberry.
I don’t know how to effectively communicate these issues right now. It feels like there are real barriers here against humans. :(
Sometimes, quality is more important than speed.
Not that any corpo developers would understand that, since their goal is to pump out as much shit as possible as fast as possible, so they can sell more shit.
I think you’re missing my point, but if you’re gonna brand me as “corpo” without giving me a chance, then hope you have a nice day, homie. I’m not the enemy here.
I’d like to address your technical argument! I’m glad to hear you’re a study of the ethics and philosophy around this because it seems like you’re intentionally sidelining those aspects of the situation, so I want to address your technical points in a way that says, “But that’s not really the whole conversation.”
In general, the best thing to do in a video game is rarely, if ever, how we should approach real-world problems; those are very controlled situations that don’t cover all side-effects or consequences. I see you’re assuming a min/max strategy but that’s also not always why we play games.
This also undersells how interrelated reading and writing are. Writing makes you a better, more critical reader, and vice versa. The converse is also true. Giving up on one of the two means being worse at both reading and writing. That puts us on a bad computer literacy trajectory as a whole.
In my view it’s different because IDE development has rather been toward, “Programmer write less boilerplate”. Automating the infrastructure like getters/setters, integrating library management, version control, so that the programmer can focus on the quality of the “tangible” result of the code. With LLM tooling, no engagement is actually necessary. I’m not saying it’s better or worse, but that it’s fundamentally different to that IDE trend you suggested.
This is the one I come closest to agreeing with, however there’s a sticking point – even if we silo the discussion to games it’s still bad for the industry as a whole. You said yourself that you’d be able to satisfyingly monetize an LLM-assisted game in 2-3 months if you got laid off, so that signals to me that throughput is the thing you find most valuable or important about this process. What that will do is make the shovelware and malware start to look closer and closer to a quality product and make the space more dangerous to be in.
If a version is bad, you can patch it. But why would you when you can generate an entirely new game with the same level of effort? If you only need 1000 people to buy software to offset the cost, why fix it at all? Not everyone will take that stance, but many will, and that will start to make it okay for everyone. It happens even now that games release early in a buggy, incomplete state with zero hope of a fix, so I’m not sure how you can responsibly gloss over this.
Games might not be banking apps, but some games process payment information or are very invasive pieces of software and should still be scrutinized for their security practices.
In short, “why wouldn’t you?” has many answers even leaving out environmental and ethical concerns, but I’m trying to keep this response small enough to meaningfully respond to if you want.
I appreciate the well written reply.
For the ethics, I personally think these things should be nationalized / universalized, given that it runs on mass theft. That’s Meta’s argument: we stole so much, you can’t hold us legally accountable. Sure thing bud: let’s get half your profit then?
For the environmental, to me, it’s akin to driving a car. It’s very socially acceptable to do that in most places in the US. Does it contribute to worsening environment and have long term impacts? Yup, sure does. I have 1 car for our family. We used to have 2. We made a positive step, but we couldn’t lower to one. Does it make me a bad person? I don’t think so. Proportionally, I don’t think my AI usage is that impactful compared to Meta/Amazon/big tech company. Is it still bad? Yup, sure is.
That’s the reason I sidestep those, along with simplifying a complex topic I knew I would be flamed for.
Literacy matters. Seems like you’re trying to push on “what if we all did it?” And yeah, there’s some idiocracy future potential. However, I don’t think it’s addressing the issue. I mentioned John Henry because that’s partially what his story is about: man v. machine. People likely wrote faster with typewriters, and I bet people claimed literacy would go down with them. Maybe? But the reading is really essential in this process. That’s my claim.
I’ll grant there’s a degree problem with the IDE analogy, but I still think it’s a technological progression that makes rational sense.
I wouldn’t make shovelware. Many people will. I’ve said to friends I expect a Cambrian explosion. Life in all forms, good or bad. We’ll get apps that could never have been made and the worst malware to exist. It’s a Pandora’s box. The internet made us closer and brought out a lot of evil in the world. To me, internet & AI are tools, not inherently good or bad. (Philosophically, if there was no thieving of data)
I think you’re making reasonable points. I also think most people have not reconciled “it’s faster to read than to type.” It’s a strange inversion in AI workflows that comes from the mechanics at play.
It’s not that we’re close. We’re miles apart. If we’re in a capitalistic hellscape (which, I think there’s maybe some evidence for, given on friendly convo :) ), then my competitor will beat me on speed. Why would I move a bunch of product on a hand cart vs. a train?
Thanks for the human interaction, by the way. I appreciate you engaging in earnest.
Thanks for your effort in replying too, I cherish these. Agree on the ethics and environmental points, I have nothing to add there!
I do feel that I understand where you’re coming from more now. It seemed like the focus of the argument before was just about the capacity of the tool (which is very interesting, I meant to ask for the numbers and forgot, thanks), so I’m glad we can get more nuanced.
Literacy matters! I struggled to keep this topic limited to video games, I did somewhat disregard the John Henry reference for that reason. I still don’t understand the claim here, though – are you saying that reading is more essential than writing for the purpose of literacy? It’d help me to reword that if you could.
I like the Cambrian explosion allegory because we saw something similar with the printing press and typewriter, and we certainly will here too. I’m coming from a much less realistic point-of-view than I’m sure you are as someone who works in the industry – Pandora’s Box is already open, we can’t hope to ban it or anything similar.
Be well, internet homie.