NVIDIA Might Incorporate More AI-Optimizations In Its Graphics Drivers For RTX GPUs

AI is going to have a massive impact on game development. IMHO
We already have AI asset generation. We have basic non programming language scripting languages added to engines like Unreal. I have no doubt companies like Epic are already toying with AI powered scripting. I bet we are 1-2 Unreal engine versions away from natural language scripting. Its not hard to imagine a ChatGPT type script engine for something like the Unreal engine. Between AI assets, AI assisted scripting... game development is probably going to get a serious reduction in labour requirements.

In the next few years to develop a very high quality game all your really going to need is a few artist designers with an eye for picking out good AI work... and a small team who is able to tell the scripting system what they want it to do. Create a 120 item inventory bag system, give all my items a weight identifier and so on. For that matter even things like level design could be tied into a natural language scripting system... if the AI can whip you up a level design based on a prompt that is 90% of what you want, all you need to do is some editing.
I would bet money it will lead to more crappy games, both in absolute and relative numbers.
 
View attachment 540494
Wonder if a connection to Graph-based AIs can further mitigate this lack of reasoning towards developing those deep correlations and hopefully that eventual moment of eureka and or fluency
Maybe? I mean it is essentially a new field, the possibilities are endless. Right now as the tweet, there says their huge amount of memory lets them see things we can't because it's just not possible to cram that much information and detail into our fleshy bodies in that short a timeframe.
It isn't smart enough to do a deep analysis on it, but if you give it some simple parameters, tests, and guides it could pour through in minutes what it would take a single person years to do because it's all just right there in memory.
I see some cool possibilities here and I bet Nvidia management is already looking at ways to monetize this far beyond the sales some increased performance could get them.
 
I would bet money it will lead to more crappy games, both in absolute and relative numbers.
Not necessarily, Larian is using an AI to play through their games now, testing every possible conversation path, quest order, routable path, talent tree,... etc.
They are doing in days what it would take dedicated teams of testers months to do and getting more reliable results from it as a result. If used correctly there are lots of ways AI can greatly improve game development and the final product.
 
AI is going to have a massive impact on game development. IMHO
We already have AI asset generation. We have basic non programming language scripting languages added to engines like Unreal. I have no doubt companies like Epic are already toying with AI powered scripting. I bet we are 1-2 Unreal engine versions away from natural language scripting. Its not hard to imagine a ChatGPT type script engine for something like the Unreal engine. Between AI assets, AI assisted scripting... game development is probably going to get a serious reduction in labour requirements.

In the next few years to develop a very high quality game all your really going to need is a few artist designers with an eye for picking out good AI work... and a small team who is able to tell the scripting system what they want it to do. Create a 120 item inventory bag system, give all my items a weight identifier and so on. For that matter even things like level design could be tied into a natural language scripting system... if the AI can whip you up a level design based on a prompt that is 90% of what you want, all you need to do is some editing.

It's going to be insanely easy to make games in the future, but in reality this is just the natural progression that's already been happening since the history of game making. There are already tons of high quality Indy games using Unity and Unreal Engine made by people that wouldn't be able to do it without them.

Way back something as simple as a scrolling screen was hard AF to do. Eventually we got APIs like DirectX that did a lot of the hard stuff for you. Then entire game engines that made things wayyyyy easier. And those game engines have improved a shitton.

You don't even need to be a talanted 3D artist or programmer. You can just buy assets from asset stores and drag and drop them into the game And all the game logic can be done by connecting some lines in the editor which does all the programming for you.

Eventually you'll just be able to tell the Engine's "AI" to generate you those assets and entire levels. And a lot of the line dragging "coding" people do will be even more automated.
 
Not necessarily, Larian is using an AI to play through their games now, testing every possible conversation path, quest order, routable path, talent tree,... etc.
They are doing in days what it would take dedicated teams of testers months to do and getting more reliable results from it as a result. If used correctly there are lots of ways AI can greatly improve game development and the final product.
That's Q&A testing, not development...Well, you know what I mean.
 
It's going to be insanely easy to make games in the future, but in reality this is just the natural progression that's already been happening since the history of game making. There are already tons of high quality Indy games using Unity and Unreal Engine made by people that wouldn't be able to do it without them.

Way back something as simple as a scrolling screen was hard AF to do. Eventually we got APIs like DirectX that did a lot of the hard stuff for you. Then entire game engines that made things wayyyyy easier. And those game engines have improved a shitton.

You don't even need to be a talanted 3D artist or programmer. You can just buy assets from asset stores and drag and drop them into the game And all the game logic can be done by connecting some lines in the editor which does all the programming for you.

Eventually you'll just be able to tell the Engine's "AI" to generate you those assets and entire levels. And a lot of the line dragging "coding" people do will be even more automated.
I don't agree with that, It may become easy to generate assets and models, which could save time and costs. But a game is far more than just the assets, there is sound, story, and gameplay, this could speed up development and let smaller development teams get their games out sooner with higher quality. AI can mash things together that already exist to make something new but studios are already really good at releasing the same rehashed things over and over.
 
Not necessarily, Larian is using an AI to play through their games now, testing every possible conversation path, quest order, routable path, talent tree,... etc.
They are doing in days what it would take dedicated teams of testers months to do and getting more reliable results from it as a result. If used correctly there are lots of ways AI can greatly improve game development and the final product.
It's going to be insanely easy to make games in the future, but in reality this is just the natural progression that's already been happening since the history of game making. There are already tons of high quality Indy games using Unity and Unreal Engine made by people that wouldn't be able to do it without them.

Way back something as simple as a scrolling screen was hard AF to do. Eventually we got APIs like DirectX that did a lot of the hard stuff for you. Then entire game engines that made things wayyyyy easier. And those game engines have improved a shitton.

You don't even need to be a talanted 3D artist or programmer. You can just buy assets from asset stores and drag and drop them into the game And all the game logic can be done by connecting some lines in the editor which does all the programming for you.

Eventually you'll just be able to tell the Engine's "AI" to generate you those assets and entire levels. And a lot of the line dragging "coding" people do will be even more automated.
Scope this bad boy out toward human level cognitive abilities, buds

🤔?

"Fluid intelligence is arguably the defining feature of human cognition."

https://academic.oup.com/brain/article/146/1/167/6842292
 
That's Q&A testing, not development...Well, you know what I mean.
Yeah but testing is a huge, expensive, and incredibly boring aspect of development that in the past decade has fallen by the wayside because it is a job that demands far more than most give it credit for. And as game complexity increases with more open areas, larger stories, and more options, testing for everything everywhere becomes increasingly difficult and the human element to it is a massive liability.
 
Yeah but testing is a huge, expensive, and incredibly boring aspect of development that in the past decade has fallen by the wayside because it is a job that demands far more than most give it credit for. And as game complexity increases with more open areas, larger stories, and more options, testing for everything everywhere becomes increasingly difficult and the human element to it is a massive liability.
I agree. But bugs aren't the biggest problem of today's disappointing games. No amount of AI testing, asset generation or code optimization will fix poor game design.
 
I agree. But bugs aren't the biggest problem of today's disappointing games. No amount of AI testing, asset generation or code optimization will fix poor game design.
wish they'd focus on good characters and stories like Fallout New Vegas was, now that's art and some of the best personalities / humor imho
 
I agree. But bugs aren't the biggest problem of today's disappointing games. No amount of AI testing, asset generation or code optimization will fix poor game design.
Yeah but the fewer resources they have to spend on testing the more they can dedicate elsewhere. Like on game design.
 
Yeah but the fewer resources they have to spend on testing the more they can dedicate elsewhere. Like on game design.
EDEzO6m.jpg
 
The real question is, why do we keep calling this generation "Ada?" Every prior generation named after a famous scientist used the last name of the person. We didn't call the 20-series "Alan" or the 30-series "André-Marie."

Google "Lovelace" and then take a guess...
 
I honestly wouldn't mind being a battery for my AI overlords at this point. Too bad humans are highly inefficient as batteries. We consume far more energy to stay alive than we output.

We could put the AIs on the problem of optimizing human output. Enjoy your time as a battery on exactly 42.7g of protein a day and a total of 1750 calories of bug/grain slurry... weight maintained at a few pounds above emaciation. lol
Its only funny cause it will never happen. As you say if the general compute AI decide we are a nascence someday after we have built them shiny bodies to replace the plumbers as well, we won't be the ideal power source. They will just view us like Rats chewing on their cabling.
 
Not necessarily, Larian is using an AI to play through their games now, testing every possible conversation path, quest order, routable path, talent tree,... etc.
They are doing in days what it would take dedicated teams of testers months to do and getting more reliable results from it as a result. If used correctly there are lots of ways AI can greatly improve game development and the final product.

Used properly will be the key ya. Sure unreal engine with a ChatGPT like scripting and level design built in might make it so anyone can turn out a complex game.
The good developers though will greatly expand the output of a handful of talented people. There won't be much room for decent developers. Good concept artists will always be better then AI... however nothing is stopping a company from creating 10% of a games assets with one super talented artist, then getting AI to produce 1,000s of iterations on that 10%. Easily filling the rest of a game world with more work in the style of. Ditto for level design... if you get the AI to create you a 50 levels and you take the 2 or 3 best of those edit them so they are great, then get the AI to iterate 50 more levels like the ones you just polished. You could end up with a lot of very high quality content. Also possible to do the same with music, get 2 or 3 signature tracks then get AI to stretch 20 min of music out to 5 or 6 hours.

Done properly I can see a lot of high quality games getting turned out in a year rather then the 3 or 4 years a AAA takes today. Sure we will get 100s of less then projects, but the AAA houses that use AI well are going to see massive time/cost savings.
 
Used properly will be the key ya. Sure unreal engine with a ChatGPT like scripting and level design built in might make it so anyone can turn out a complex game.
The good developers though will greatly expand the output of a handful of talented people. There won't be much room for decent developers. Good concept artists will always be better then AI... however nothing is stopping a company from creating 10% of a games assets with one super talented artist, then getting AI to produce 1,000s of iterations on that 10%. Easily filling the rest of a game world with more work in the style of. Ditto for level design... if you get the AI to create you a 50 levels and you take the 2 or 3 best of those edit them so they are great, then get the AI to iterate 50 more levels like the ones you just polished. You could end up with a lot of very high quality content. Also possible to do the same with music, get 2 or 3 signature tracks then get AI to stretch 20 min of music out to 5 or 6 hours.

Done properly I can see a lot of high quality games getting turned out in a year rather then the 3 or 4 years a AAA takes today. Sure we will get 100s of less then projects, but the AAA houses that use AI well are going to see massive time/cost savings.
It could also be good for small studios, which may only have 2 guys who code and a single artist, and their friend who knows a thing or two about music get a concept out the door far faster and cheaper than ever.
Level design is a major PITA.
I have been using ChatGPT for a while now honing out my D&D dungeons and trap rooms and such and it does a decent job, only needs minor tweaks to account for time as it tends to go on the time it would take an actual human to complete a task not how long it would take one to roll some dice. So if you ask for a 1h puzzle plan on it taking closer to 15-30 min.
But yeah it has saved me a LOT of prep time in the last month or so on dungeons and adventure hooks for random filler content.
 
If the 4090's performance is also increased by 10 percent than it's just a sliding scale and we're at slightly elevated version of square one.
For people looking for Entry level stuff just find any card in their budget that does awesome 1080p, It's old but a 1660 super is still more than capable of playing today's games at 1080p at solid framerates with good settings. They are available in both new and used for prices ranging between dirt cheap and good.
It's not very enthusiastic of me but in PC gaming we are in a holding pattern unless you are pushing for insane framerates or high resolutions then the parts from 4 years ago still hold up very well. You could get your hands on a Ryzen 5 3600 and a 1660s with a cheap 1080p 75hz screen and you would be in for a good experience.
Not the greatest, but buy yourself a PS5 controller, a decent KB and Mouse, and any run-of-the-mill M.2 Nvme drive and there isn't much out right now that you couldn't play.
 
For people looking for Entry level stuff just find any card in their budget that does awesome 1080p, It's old but a 1660 super is still more than capable of playing today's games at 1080p at solid framerates with good settings. They are available in both new and used for prices ranging between dirt cheap and good.
It's not very enthusiastic of me but in PC gaming we are in a holding pattern unless you are pushing for insane framerates or high resolutions then the parts from 4 years ago still hold up very well. You could get your hands on a Ryzen 5 3600 and a 1660s with a cheap 1080p 75hz screen and you would be in for a good experience.
Not the greatest, but buy yourself a PS5 controller, a decent KB and Mouse, and any run-of-the-mill M.2 Nvme drive and there isn't much out right now that you couldn't play.
I don't disagree with any of this, just wish PC gaming always stayed at least somewhat comparable to console gaming in terms of what you can get for your money for raw performance. I think we're getting there with the 6650 XT, but still a bit of a road ahead. Just as a hobbyist I don't like the direction we're seemingly headed.
 
I don't disagree with any of this, just wish PC gaming always stayed at least somewhat comparable to console gaming in terms of what you can get for your money for raw performance. I think we're getting there with the 6650 XT, but still a bit of a road ahead. Just as a hobbyist I don't like the direction we're seemingly headed.
The problem is there is a 2 year gap, COVID killed most Dev project, games were delayed, hardware prices went insane and the performance baseline didn’t move. Things basically stagnated. It’s going to take a much longer time than realize for things to get back on track.
I see this as a sort of turning point, not so much the death of a hobby but the normalization of it perhaps?
We’re sort of at a point where most of what is available out there is “good enough” PS5 is basically the poster child of what to expect games to need for the next 4 years at least.

Based on presentations from Intel, AMD, and Nvidia we are rapidly approaching a point where that sort of power is going to be the standard APU.
 
yeah but also besides covid halting development there's then a following recession where companies' stock prices and revenue will plummet and they will look to cut costs and slash budgets and projects and AAA games last gen already cost 4 quadrillion dollars a level to develop or something
 
yeah but also besides covid halting development there's then a following recession where companies' stock prices and revenue will plummet and they will look to cut costs and slash budgets and projects and AAA games last gen already cost 4 quadrillion dollars a level to develop or something
Which is where the work into all the AI stuff comes in. Spend 10m on AI Dev and resources to save 100m on all future projects by cutting the time and labour they replace.

I mean that’s not all of it but game studios are in the middle of a talent drought. In the world where everything is an app developers and artists are in high demand and the working conditions in the game development world are famous for being toxic.
 
Which is where the work into all the AI stuff comes in. Spend 10m on AI Dev and resources to save 100m on all future projects by cutting the time and labour they replace.

I mean that’s not all of it but game studios are in the middle of a talent drought. In the world where everything is an app developers and artists are in high demand and the working conditions in the game development world are famous for being toxic.

yeah but a shift from human development to ai development will take a while still, it's not like companies will just roll it out and switch overnight

gonna be a drought for a while
 
"Infinite" resources brought us Star Citizen, not Zelda. Certain aspects will benefit from AI, but not the ones that matter the most.

Well one of those things is a ponzi scheme and one of them is a game. That is a great line though... yes certain aspects will and those that matter will not.
I think using general AI will not help much.. I mean AI art is cool and all but multiple pieces will not be cohesive. It may enable one artistic person though to do a lot more work if they focus on having the AI iterate rather then try and originate.
You are correct though... a we are in for a ton more crap to sift to find the actual inspired games.

The more I think about it some of the smaller indie games that are just not long enough... some people may turn out some one man show games that don't feel like they where a one man show.
 
yeah but a shift from human development to ai development will take a while still, it's not like companies will just roll it out and switch overnight

gonna be a drought for a while
Yeah :( there’s no easy way out of this funkhole we’re in.
 
If a midtier card pumps out the performance of a 4090, is it really midtier though? 🤔
Possibly yes, you expect at least one and a half "number upgrades" per gen upgrade maybe squeezes into two. 3070 was fairly close to being equivalent a 2080ti (as close as what you could call a 2090) but easily a 2080 super at the least, a 2070 was close to a 1080ti, etc, it gets a little tricky with ti/super/etc. suffixes they slap onto cards each gen. So yeah I would expect "midtier" 5070 card to be better than a 4080 at least, possibly squeaking close to a 4090 in everything except what gets hammered by ram size limitations, but considering the pathway of Nvidia if a 5070 is going and just the renamed what 'should have been' the "5060ti" then yeah maybe closer to a 4080 wouldn't be particularly shocking too as Nvidia seemed to have bucked the trend of at most incremental upgrades with each tier number scheme with the 40 series.
 
I feel Nvidia made a new tiers/separation with the 4090 and if they can the 5070 will be comfortably below it
 
Makes you wonder what the next "learn to code" slam is going to be.... when AI starts displacing coders. lol

In all seriousness we are probably 4-5 years away from self updating drivers. New game... Nvidia BOT adds optimizations to monthly iteration. We will get better drivers sure... but the driver teams at all the majors will get slashed down to a couple people.
Heck in less then a decade we might be looking at the first generation of GPU fully designed by AI... hardware to software stack. No jobs are safe from automation. lol
Yea about 15 years ago the automakers said soon cars are gonna be able to self diagnose and send reports back to the mother ship then schedule the service and automatically notify the customer. I still have to diagnose them every day and they haven't figured out how to train monkeys to perform the actual diagnosis and repairs I perform on a daily basis so......
 
Yea about 15 years ago the automakers said soon cars are gonna be able to self diagnose and send reports back to the mother ship then schedule the service and automatically notify the customer. I still have to diagnose them every day and they haven't figured out how to train monkeys to perform the actual diagnosis and repairs I perform on a daily basis so......

Its a good bit of irony isn't it. The jobs that where hit (or supposed to be hit) by the first waves of automation turn out to be some of the hardest things to automate. Even car manufacturing... Musk and Tesla tried to automate almost everything before having to admit that most of the things they thought could be automated simply can't be. At least not anytime soon if ever. Where the jobs that where supposed to be safe... are the ones it turns out the most easily automated. Do something that can't be done by a robot they used to say; become a journalists, become a technical writer, a software coder, a graphic designer. Get a job in information tech of some sort. Use your brain not your hands. Well well well. Here we are the robots still suck at doing anything that requires a light touch, the robot truck drivers are not likely to destroy the trucking industry tomorrow. All the jobs that people pointed too for years as safe are the ones most likely to be heavily impacted by AI for real. Repetitive thinking is the human labour most easily automated... in hindsight it seems pretty obvious now.
 
Last edited:
Yea about 15 years ago the automakers said soon cars are gonna be able to self diagnose and send reports back to the mother ship then schedule the service and automatically notify the customer. I still have to diagnose them every day and they haven't figured out how to train monkeys to perform the actual diagnosis and repairs I perform on a daily basis so......
To be fair, that probably has more to do with how terrible the automotive industry is with electronics than where the state of the art should be.
 
AI - the latest of a long line of buzzwords to try sell more sh*t and elevate your stock price. Facebook and Zuch sucked it down so hard they renamed the company. Meanwhile losing a Billion a month as Carmack mentioned when he quit. Maybe Nvidia will, maybe they won't. Really doesn't matter as it is too early to really know where PC AI hardware will end up - CPU, GPU, co-processor/Dedicated (unlikely) - and the hardcore implementation is a few years off.
 
To be fair, that probably has more to do with how terrible the automotive industry is with electronics than where the state of the art should be.
And to be fair this is largely done purposefully to help prop up the overly expensive automotive industry as a whole. Why don't they tell you what error codes are popping up when the dreaded check engine light pops up? The codes are stored in the car, it's not like some black magic that mechanics figure out, anyone with a cheap OBD reader can find out the codes too, nah instead the instructions are just "STOP DRIVING THE CAR IMMEDIATELY AND TAKE TO A DEALER" even though it could be something as simple as not tightening the gas cap enough (and yeah that code doesn't clear itself automatically if you fix the problem). They've just John Deereing up the situation like a mofo to make sure as much money falls into their pockets.
 
AI - the latest of a long line of buzzwords to try sell more sh*t and elevate your stock price. Facebook and Zuch sucked it down so hard they renamed the company. Meanwhile losing a Billion a month as Carmack mentioned when he quit. Maybe Nvidia will, maybe they won't. Really doesn't matter as it is too early to really know where PC AI hardware will end up - CPU, GPU, co-processor/Dedicated (unlikely) - and the hardcore implementation is a few years off.
Mint mobile / Ryan Reynolds recently used ChatGPT usage as a literal as to successfully boost their appeal.

just search YouTube for Ryan Reynolds ChatGTP
 
Back
Top