[WOW!] Google’s GameNGen: AI breaks new ground by simulating [interactive] Doom without a game engine

[Nilgiri]

From a business standpoint, I don't think AI will replace the human workforce in its totality, as you pointed out, as it's impossible. However, it will replace the menial jobs we do at the office. I'm dealing with a client on the manufacturing side using AI, and it's a give-and-take sort of situation. However, it will complement the work we are currently doing, carrying out some grunt work.

The insurance industry, which is acting as my premium float for other businesses, is seeing rapid AI adoption, and it's paying us dividends in increased profits and margins from rating (actuarial work), underwriting, loss control, adjusting for claims, auditing, etc.

Yes actually AI can create a lot more (human) jobs in sectors too as new things (governing human time and our subjective value/preferences that dictates economics etc).... are unlocked. Other sectors will be more straightforward attrition and replacement by AI.

But human jobs are not going away, things will just shift over time.

 

[PakFactor]

Yes actually AI can create a lot more (human) jobs in sectors too as new things (governing human time and our subjective value/preferences that dictates economics etc).... are unlocked. Other sectors will be more straightforward attrition and replacement by AI.

But human jobs are not going away, things will just shift over time.

Indeed. However, I feel the UBI question will be at the forefront again. AI will wreak havoc on pay overtime, leaving us with a position with too few postings and too many applicants.

 

[MNZGamerX]

I played Doom in the 90's and this AI generated version looks and feels exactly the same and that is an astonishing feat. If AI can do this now imagine its capabilities in 5 years or 10 years time. AI only took off in 2022 and in a matter of just 2 years here we are. This technology has a lot of potential but will eventually make humans redundant.

 

[AjayGhatak]

The whole education model has to be reformed and changed to adapt to what AI will be doing this century moving forward. Humans are throwing too much resource at redundant sectors for themselves still....and throwing large resource at keeping it afloat afterwards too. I think AI will do its role in vast undercutting and resulting forced reform in the entire system. Wont be overnight, but it will work in tiers slowly.

These days, I avoid learning entire software frameworks but learn basic principles upon which it is architected. This helps me to quickly get a feel of what may or may not be possible with that framework. Rest I just autogenerate and verify. I also debate with the model a lot. In the sense, once it suggested me a particular fashion to solve a particular problem. I asked if it could be solved in a different way by relaxing a constrain it was assuming. It said it should be possible and gave me a sketch of the solution which I refined and completed my work.

Human knowledge as an explorer of concepts is becoming more important than all the nuts and bolts. Having experience in doing nuts and bolts is still very useful to refine solutions but this way makes things much more faster for me. I can complete six months of work in six days or less this way.

Also, LLMs are very good tool for sound boarding and learning new things.

 

[AjayGhatak]

I played Doom in the 90's and this AI generated version looks and feels exactly the same and that is an astonishing feat. If AI can do this now imagine its capabilities in 5 years or 10 years time. AI only took off in 2022 and in a matter of just 2 years here we are. This technology has a lot of potential but will eventually make humans redundant.

I looked at the architecture of the model given. To be honest it is not a massive leap. It is very similar to video generator from image models but instead of language embedding it is embedding game inputs to shape the video getting generated. An incremental improvement.

One must realize that this may or may not learn the actual game engine at all. Actually, it is very black box way of learning still now. If there were parts of level which were not visited, this will not be able to fully predict them (ofcourse). But it will fill those gaps with thematically matching output.

What will REALLY wow me is if a model can also use object/script code and game data to learn how game works and allows you to generate totally new levels. That means that model is no only predicting output (by learning "shape" of the data) based on input, it has also incorporated all the rules and their implications.

Now, in my wildest dreams, I dream of next generation processors that are devoid of instructions but can be retrained to accomodate different types of transformers and other models. Basically, computation changing from a series of instructions to a network of connects and thresholds that can be adjusted (like an analog FPGA) and that becomes the bedrock of implementing intelligent solutions.

 

[MNZGamerX]

I looked at the architecture of the model given. To be honest it is not a massive leap. It is very similar to video generator from image models but instead of language embedding it is embedding game inputs to shape the video getting generated. An incremental improvement.

One must realize that this may or may not learn the actual game engine at all. Actually, it is very black box way of learning still now. If there were parts of level which were not visited, this will not be able to fully predict them (ofcourse). But it will fill those gaps with thematically matching output.

What will REALLY wow me is if a model can also use object/script code and game data to learn how game works and allows you to generate totally new levels. That means that model is no only predicting output (by learning "shape" of the data) based on input, it has also incorporated all the rules and their implications.

Now, in my wildest dreams, I dream of next generation processors that are devoid of instructions but can be retrained to accomodate different types of transformers and other models. Basically, computation changing from a series of instructions to a network of connects and thresholds that can be adjusted (like an analog FPGA) and that becomes the bedrock of implementing intelligent solutions.

Quantum computers may provide that level of computational power and more

 

[Hamartia Antidote]

I played Doom in the 90's and this AI generated version looks and feels exactly the same and that is an astonishing feat. If AI can do this now imagine its capabilities in 5 years or 10 years time. AI only took off in 2022 and in a matter of just 2 years here we are. This technology has a lot of potential but will eventually make humans redundant.

This video happens to have the actual Doom level shown in the footage above and it is hard to tell the difference