Support NeoGAF

[IbizaPocholo]

Sony’s Shuhei Yoshida says AI will change the nature of learning for developers | VGC

“People might not even need to learn programming any more,” according to the PlayStation executive…

www.videogameschronicle.com

In an interview with The Guardian, Yoshida was asked what he thinks about developers’ concerns that AI could replace human effort in disciplines like art, music and coding.

Yoshida said developers will need to learn new skills to effectively use AI, which he believes will ultimately have a positive impact on game development.

“It is a tool. Someone has to use the tool,” he said. “AI can produce very strange things, as you must have seen. You really have to be able to use the tool well. AI will change the nature of learning for game developers, but in the end development will be more efficient, and more beautiful things will be made by people.

“People might not even need to learn programming any more, if they have learned how to use these tools of the future. The creativity is more important, the direction, how you envision what you want.”

Yoshida joined the PlayStation project in 1993, a year before the release of the original console, and went on to become the president of Sony Computer Entertainment Worldwide Studios.

He currently serves as head of SIE’s Independent Developer Initiative, where his role is to nurture external, smaller independent studios.

Yoshida gave The Guardian an example of how AI is being used by smaller teams to help them create projects that might otherwise be out of their capability.

“I was going through 15 pitches in a competition for indies in Japan just this morning, and one of them had amazing beautiful graphics made by a small team of students,” he said.

“They said that they used Midjourney, the AI art generator, to create the art. That is powerful, that a small number of young people can create an amazing looking game. In the future, AI could develop interesting animations, behaviours, even do debug for your program.”

 

[feynoob]

I passed my college classes thanks to chatgpt.

The instant answer with detailed explanation helps a lot, especially when you are struggling with the questions.

 

[Heimdall_Xtreme]

PLEASE!!!!

COME BACK!!! SUSHI!!!

WE MISS YOU!!!

 

[Danjin44]

 

[Heimdall_Xtreme]

yahoooo???

 

[CGNoire]

"Please be Excited about our ForthComing Job Cuts"

 

[{insert code here}]

In my experience, chatgpt has made most programmers worse at completing tasks. Instead of using it as a tool, most use it as a crutch.

I think one possibility is AI will ultimately turn the few who know how to use it properly into God tier engineers but everyone else will be worse off.

 

[intheinbetween]

In my experience, chatgpt has made most programmers worse at completing tasks. Instead of using it as a tool, most use it as a crutch.

I think one possibility is AI will ultimately turn the few who know how to use it properly into God tier engineers but everyone else will be worse off.

ChatGPT 3.5 (or 4) will fix most of the issues on that regard, however I'm not sure if Codex would be a better approach for prorgamming than ChatGPT 3.5 since OpenAI recommends moving from Codex to ChatGPT 3.5

 

[AV]

"Will"? It already is.

I work in software development and we're already using ChatGPT for easy and concise instructions and sample code to solve very niche and sometimes complex problems. It's absolutely insane how good it is and how much time it saves you trawling decade old Stack and Github pages for outdated answers. You can even feed it error messages and it'll tell you why it was mistaken and update its own code.

Not perfect, far from, and still requires an understanding of the code you're copying and pasting. But it's a remarkable tool.

 

[RoboFu]

Here I thought Unreals blueprints where dumbed down a lot lol.

Creatives will actually be the first to go. Art assets and music are simple things to replace in games. Asking a ai for 6 different types of repeatable rock textures is an easy task.

 

[TheThreadsThatBindUs]

"Please be Excited about our ForthComing Job Cuts"

Nah, it's more like "Please be excited for a significant reduction in development cost and time"... meaning top-end blockbuster AAA games developed in a 2-3 yr cycle again, and indie devs on UE5 will be able to put out games with AAA-level graphics and asset scope using AI for procedural generation.

Games will be able to be bigger and have more depth than ever before.

Look at titles like No Man's Sky in its current state. It's a fantastic example of how procedural generation tech has allowed a small indie dev to make a game with a scope far exceeding the efforts of some of the biggest AAA studios. The game has a ridiculous amount of depth today too.

 

[Fuz]

People might not even need to learn programming any more

This has "disaster" written all over it.

 

[Fake]

Devs: No more crunch
AI: OK
Devs: No u

 

[sendit]

This has "disaster" written all over it.

How so? Before high level languages, developers had to write in assembly. This is all but abstracted now. Layers of abstraction will just continue to increase, and the final output is really all that matters to the end user.

 

[DemonGamerTTT]

Will? What year is this a again?

We're already doing that, anyone that works with software development needs to start changing how they work now...

Those who don't will be unemployed 6 months to 1 year from now.

 

[EDMIX]

Will? What year is this a again?

Is already doing that, anyone that works with software development needs to start changing how they work now...

Those who don't will be unemployed 6 months to 1 year from now.

Pretty much this 1000x

This is an unavoidable thing. To not use it, is literally to stay in the past as if knowing less is going to help or something. I get many want to add some moral thing to this like jobs and shit, but what company is taking the L to not use Ai, to watch their enemies feast on them? Its why I never get this whole "Why do Walmart do dis doe" ideology, cause what they don't do, Amazon will, Target will etc. It is a kill or be killed type of thing and NOT using Ai or automation only serves to hurt the business thinking they can some how force the whole industry to follow them or something. So I can't fault Sony or MS for jumping into Ai and finding ways to use it, their enemies won't just sleep on Ai or something.

 

[sachos]

It will affect EVERY part of game dev. And it will keep getting better, even now GPT-4 is only limited to 8k tokens, the 32k tokens version is already enough to pass whole documentation to it and have figure out the solution to a problem from it and there are papers coming out that propose methods for increasing context length to 1 or 2 million tokens, thats like throwing in 3 whole books and asking it to write a sequel. Imagine that but for an entire code base, it will be nuts.

 

[Mikey Jr.]

I was playing Trails of cold steel series.

This was a few years ago. And I thought to myself "would it ever be possible for a small series like this to get full voice acting?"

And I mean, EVERY single line of dialogue in this huge JRPG. The shopkeeper, the random dude on the street, etc. Like, don't replace the human actors from main roles, but use AI to voice all the flavour text from the NPC's.

That is what I'm excited about. That these small developers can really increase the presentation of their games by a huge amount, but the cost of that would be so little.

 

[decisions]

This has "disaster" written all over it.

The reason why I don’t believe this sentiment is that it is all based on the idea that English is somehow a perfectly accurate, effective language for describing how you want a machine to work.

But you won’t get through the first year of college thinking that. English is a terribly ambiguous, verbose language, and that is why we write code.

 

[AtomicStarving]

Using AI to fix bugs

 

[{insert code here}]

ChatGPT 3.5 (or 4) will fix most of the issues on that regard, however I'm not sure if Codex would be a better approach for prorgamming than ChatGPT 3.5 since OpenAI recommends moving from Codex to ChatGPT 3.5

If you've seen what I've seen then I don't think you would be this optimistic. You can prompt it, hey how can I solve problem A. Or you can prompt it with, hey solve problem A. I see people almost always doing the latter and this is how it's made people worse at tasks because they aren't learning or doing anything. They're copy pasting. With the first prompt you're still doing work because you've only reduced gpt into a hyper focused form of Google search. This is how it will make already competent engineers into beasts because we can find the exact information we need instantly. No more digging through endless pages of documentation searching for one small function or parameter.

 

[CGNoire]

Nah, it's more like "Please be excited for a significant reduction in development cost and time"... meaning top-end blockbuster AAA games developed in a 2-3 yr cycle again, and indie devs on UE5 will be able to put out games with AAA-level graphics and asset scope using AI for procedural generation.

Games will be able to be bigger and have more depth than ever before.

Look at titles like No Man's Sky in its current state. It's a fantastic example of how procedural generation tech has allowed a small indie dev to make a game with a scope far exceeding the efforts of some of the biggest AAA studios. The game has a ridiculous amount of depth today too.

Not Worth what where losing. Im the biggest graphics whore and that shit aint worth the price.

 

[lyan]

If you've seen what I've seen then I don't think you would be this optimistic. You can prompt it, hey how can I solve problem A. Or you can prompt it with, hey solve problem A. I see people almost always doing the latter and this is how it's made people worse at tasks because they aren't learning or doing anything. They're copy pasting. With the first prompt you're still doing work because you've only reduced gpt into a hyper focused form of Google search. This is how it will make already competent engineers into beasts because we can find the exact information we need instantly. No more digging through endless pages of documentation searching for one small function or parameter.

For current working generation, in future where production rate expects copy pasting few will have the luxury to go through the ground work to become the beast, and humanity will be left with fewer competent engineers, until the AI can self evolve and replaces us entirely at least.

 

[Varteras]

I passed my college classes thanks to chatgpt.

The instant answer with detailed explanation helps a lot, especially when you are struggling with the questions.

A guy I know who works for the government in California said he used an AI to literally do his job for him. Came up with a huge, well-written document for a legal matter. Said in that moment he realized how close he is to not having a job anymore.

 

[sinnergy]

And he is right .

 

[Kumomeme]

A.I gonna be like smartphone

where it mark the generation differences.

 

[NahaNago]

Nah, it's more like "Please be excited for a significant reduction in development cost and time"... meaning top-end blockbuster AAA games developed in a 2-3 yr cycle again, and indie devs on UE5 will be able to put out games with AAA-level graphics and asset scope using AI for procedural generation.

Games will be able to be bigger and have more depth than ever before.

This is why I don't think AAA games will go back to the 2-3 yr development cycle. They are going to figure out how to effectively use ai and then increase what they want to create in the game.

Take Starfield for example. Why create 1,000 planet levels? and the simple answer is because they were now able to.

 

[TheThreadsThatBindUs]

Not Worth what where losing. Im the biggest graphics whore and that shit aint worth the price.

What are you losing?

 

[TheThreadsThatBindUs]

This is why I don't think AAA games will go back to the 2-3 yr development cycle. They are going to figure out how to effectively use ai and then increase what they want to create in the game.

Take Starfield for example. Why create 1,000 planet levels? and the simple answer is because they were now able to.

Nah... Publishers want to be able to account for regular revenue in their annual financials. With games being more expensive and taking longer to make, with snazzy AI tools that increase productivity massively, they aren't going to balloon game scope and dev time.

Instead, they'll push out more games for the same studio overhead costs, since they will generate more sales revenue in aggregate.

Why ship 1x ridiculously large game in 10 yrs, when you can ship 3x reasonably sized games for the same studio size in the same time period and spread the risk over 3x projects? What you propose doesn't make a lot of sense for devs.

 

[DareDaniel]

 

[GymWolf]

This has "disaster" written all over it.

It's a double edged sword.

Knowing how to program doesn't automatically mean that you have great ideas for games, and who knows how many people around the world have great ideas for videogames but they can't do jack shit because they don't know how to program.

 

[ResilientBanana]

This is incredible.

 

[SteelCurtain 59]

AI will be a massive net positive for gaming in both quality and quantity of games, it will definitely cost people some jobs though, which is obviously shitty.

 

[{insert code here}]

For current working generation, in future where production rate expects copy pasting few will have the luxury to go through the ground work to become the beast, and humanity will be left with fewer competent engineers, until the AI can self evolve and replaces us entirely at least.

Okay yeah, then I misinterpreted your original quote because we basically agree in our conclusions that there will be fewer and fewer competent engineers over time. AI is likely going to fill that void with very small teams of engineers working in concert or backed by dozens of AI entities.

 

[DEAD ORIGINAL]

Using AI to fix bugs

This imagine AI fixing all bugs prior to launch play testing doing All the tedious jobs allowing the Devs to be just creative

 

[NahaNago]

Nah... Publishers want to be able to account for regular revenue in their annual financials. With games being more expensive and taking longer to make, with snazzy AI tools that increase productivity massively, they aren't going to balloon game scope and dev time.

Instead, they'll push out more games for the same studio overhead costs, since they will generate more sales revenue in aggregate.

Why ship 1x ridiculously large game in 10 yrs, when you can ship 3x reasonably sized games for the same studio size in the same time period and spread the risk over 3x projects? What you propose doesn't make a lot of sense for devs.

With every upgrade in tools or hardware for games, studios seem to be making games more detailed and larger so why wouldnt they do the same thing with the benefits from ai.

I agree that they they should make 3 and maybe even 4 games in 10 years it is just a matter of what the studio wants to make. The question though is will they. I'd personally like more smaller AAA games.

 

[Fuz]

It's a double edged sword.

Knowing how to program doesn't automatically mean that you have great ideas for games, and who knows how many people around the world have great ideas for videogames but they can't do jack shit because they don't know how to program.

Not really thinking about videogames there.

 

[TheThreadsThatBindUs]

With every upgrade in tools or hardware for games, studios seem to be making games more detailed and larger so why wouldnt they do the same thing with the benefits from ai.

I agree that they they should make 3 and maybe even 4 games in 10 years it is just a matter of what the studio wants to make. The question though is will they. I'd personally like more smaller AAA games.

I've given you the reason. Because they're bound by the whims of the publishers who pay the bills. And pubs don't care about game size or detail. They care about their bottom line. It's why we get so many AAA rushed out the door before they're ready; only to be patched later. Publishers control how long devs can burn through development capital on a project.

$500m 10yr dev cycles are not something publishers will stomach. But publishers will lap up the prospect of being able to develop a full-fat, full-featured CoD game every two years again. That's over a billion dollars every two years they get to account for in their revenue. That's the type of shit that makes investors giddy like schoolchildren, and investors are who the publishers are beholden to.

 

[NahaNago]

I've given you the reason. Because they're bound by the whims of the publishers who pay the bills. And pubs don't care about game size or detail. They care about their bottom line. It's why we get so many AAA rushed out the door before they're ready; only to be patched later. Publishers control how long devs can burn through development capital on a project.

They've always been bound by the whims of publishers and yet games are still taking a lot longer to make even with all of the upgrades in video game engines and hardware. Devs are going to try and be as ambitious as possible and publishers are going to be like you need to release the game in the next 6 months.

$500m 10yr dev cycles are not something publishers will stomach. But publishers will lap up the prospect of being able to develop a full-fat, full-featured CoD game every two years again. That's over a billion dollars every two years they get to account for in their revenue. That's the type of shit that makes investors giddy like schoolchildren, and investors are who the publishers are beholden to.

I agree that publishers would want everything you are saying. I just don't think they would cut CoD development down to 2 years again. They could create a full featured cod game every 2 years now. They simply keep adding what they want to put in the game to make development even longer.

 

[Tams]

How so? Before high level languages, developers had to write in assembly. This is all but abstracted now. Layers of abstraction will just continue to increase, and the final output is really all that matters to the end user.

This.

How many coders, especially web developers, know what's actually going on with they code they write?

Excluding those who know (and ideally use) low level languages, it's probably, what? 1%? 3% at a push?