NVIDIA’s GameGAN AI Recreated PacMan! 👻
Вставка
- Опубліковано 19 чер 2020
- ❤️ Check out Weights & Biases and sign up for a free demo here: www.wandb.com/papers
The mentioned blog post is available here:
www.wandb.com/articles/visual...
📝 The paper "Learning to Simulate Dynamic Environments with GameGAN" is available here:
nv-tlabs.github.io/gameGAN/
Our paper with the neural renderer is available here:
users.cg.tuwien.ac.at/zsolnai...
🙏 We would like to thank our generous Patreon supporters who make Two Minute Papers possible:
Aleksandr Mashrabov, Alex Haro, Alex Paden, Andrew Melnychuk, Angelos Evripiotis, Benji Rabhan, Bruno Mikuš, Bryan Learn, Christian Ahlin, Daniel Hasegan, Eric Haddad, Eric Martel, Gordon Child, Javier Bustamante, Lorin Atzberger, Lukas Biewald, Michael Albrecht, Nader S., Owen Campbell-Moore, Owen Skarpness, Robin Graham, Steef, Sunil Kim, Taras Bobrovytsky, Thomas Krcmar, Torsten Reil, Tybie Fitzhugh.
More info if you would like to appear here: / twominutepapers
Meet and discuss your ideas with other Fellow Scholars on the Two Minute Papers Discord: discordapp.com/invite/hbcTJu2
Károly Zsolnai-Fehér's links:
Instagram: / twominutepapers
Twitter: / twominutepapers
Web: cg.tuwien.ac.at/~zsolnai/
#NVIDIA #GameGAN - Наука та технологія
So this explains how EA can recreate the same games so fast.
No, EA doesn't have intelligence, much less artificial intelligence.
No. EA doesn't "recreate" games. It just releases the same game with a different title and some real life data updates.
Francisco Russo that’s how sports games work, dude. Would you be happy if Madden 2021 had farm animals instead of humans playing the game? Would that be innovative enough?
@@ophello My critic is that it is immoral to sell a game EVERY YEAR for very high prices, instead of being 1 every some years, (with some regular updates), like most new games are.
As if EA invested that much money in AI. 97% of the money goes for the shareholders pocket and 3% for doing the games, let alone for AI
Soon we won't just have procedurally generated worlds, but even procedurally generated rules and gameplay. You will buy a game that will just say "We have no idea what this game will be when you play it, only that it will evolve to be fun in a way that is specific to you."
so almost like the AI Dungeon?
Ohh I like that a bit like Rimworld but freer. Also it means unlimited mod support.
While it's an interesting idea, it is not what that AI does. It doesn't generate rules, it generate code that fit the rules it deduced from the footage you have shown him. Most likely direct application I can see would be in game preservation to resurect old games whose sourcecode were lost or who used to work on hard to find / emulate consoles. Another (less moral) direct application would be to produce cheap knock-offs of popular games.
@@tetri90 Yeah, while that's true for this specific AI, what I was referring to is more the direction that a lot of AI seem to be going, and how combining a few of them will create a whole new genre of game that builds itself for you. We're a long ways off, but in the grand scheme of things, it's still "soon".
@@tetri90 Could also imagine creating a base game gameplay from a video of what the game should look like and desired inputs.
Another application we could imagine is game porting. Looking at a game gameplay and inputs, map the inputs and let a specific model translate the game from one engine to another
AI:” You know the rules, and so do I”
AI learns to make AI, two papers down the line and we'll reach the singularity.
ua-cam.com/video/ub82Xb1C8os/v-deo.html
@@craftmechanics6483 stunning.
I was about to make this comment, but you beat me to it
And it will be THE END.
ua-cam.com/video/GHMjD0Lp5DY/v-deo.html
Penta haha noce
you can see dots that were eaten being partially rendered, because there's always some residual output from the initial state of the game being displayed, as it is the most frequent state in the training data.
Two more papers and this will get better and better!
I think all it needs is more training data
Whats the quote about kids running around eating strange pills and listening to techno...
...
...
Does LSD cause lag?
Or enable ~hax~ ¿
Wish some neuroscientist would get in on some of this. It's seriously so akin to how our own minds work it's getting a bit crazy now. Can't wait for the next breakthrough.
those were in the static objects layer.
I'd like to see a comparison in CPU performance and memory consumption between AI generated programs and manually written programs.
@Axolotl Is Cool There are three different neural networks coordinating with each other to generate each frame. There's no way that's more efficient than the close to the metal original implementation of pacman that had to squeeze everything out of limited hardware. Even if it was trained perfectly so it was determining the exact rules that were used to code the original game I would still imagine it would be orders of magnitude less efficient. Could be wrong though. No benchmarking on the linked site. Performance is far from the point here, impressive regardless.
googles autoML is more efficiant than its man made counterparts. From skim reading I think it said about twice as efficiant.
@@grant5642 Well, you are wrong. In this channel you can see how AI can simulate physics being litterally orders of magnitude efficient than current simulation software.
Pacman wont be a big deal, pal...
This entire channel is most likely generated by an AI
I could entirely believe that he would just train a neural network on his voice and automate that. It probably wouldn't be that hard nowadays. :P
2 millisecond papers.
Can't unsee :OOO
The application for this in video games could be staggering, imagine an Ai that can take a look at the games you play often and enjoy the most and then take elements from those and create something personally designed to be the perfect game for you.
I love your job, too
Exactly 😂👌🏻
I job your love!
So the computer programs can write complicated computer programs now.
Time to teach a neural network to write neural networks that write neural networks!
Skynet time.
"Let's invent a thing inventor," said the thing inventor inventor after being invented by a thing inventor.
This sounds like a joke but it’s not. Skynet for sure.
@@imveryangryitsnotbutter where is that from?
@@Supreme_Lobster yeah I'm wondering too it sounds familiar. Could be an exurb1a line?
they are doing magic
this is not okay
- When a distinguished but elderly scientist states that something is possible, he is almost certainly right. When he states that something is impossible, he is very probably wrong.
- The only way of discovering the limits of the possible is to venture a little way past them into the impossible.
- Any sufficiently advanced technology is indistinguishable from magic.
I for one welcome our new AI overlords.
xavdel0 it’s Michio Kaku quote, you should write it between brackets
@nasolem Huh, true..
As a programmer, this is the first time that I'm worried about losing my job to AIs lol
You have not seen it all check this amazing OpenAI project: ua-cam.com/video/fZSFNUT6iY8/v-deo.html
i have thought about that for the entirety of the video 😅
So, I guess even "learn to code" isnt safe from being automated. ¯\_(ツ)_/¯
If this is where AI is at in 2020 I can't imagine what people will come up with by 2030
Singularity was estimated to arise around 2033. There are still traces of "2033" appearing here and there: books.google.be/books?id=xZUnDwAAQBAJ&pg=PA51&lpg=PA51&dq=singularity+2033&source=bl&ots=exdXo08Zpm&sig=ACfU3U2UtJWU8LrFPKc1BaI4GO7OIySiRw&hl=en&sa=X&ved=2ahUKEwjikO2yypDqAhVFqaQKHZVvA6Y4ChDoATADegQICRAB#v=onepage&q=singularity%202033&f=false
@@programaths that prediction depends on who you ask
@@CamKnoppMusic Exact!
It would be nice to be able to create a game just by making an animation of it (paired with desired input).
That sounds possible based on this paper but you would need hundreds of hours of animation
There are models that can generate images from a text description. I can imagine sometime could develop a system where we could write a description of a number of parts to a game (eg "And if the player picks up the key, they can unlock those two doors, which first glow with a mysterious pattern before fading away") and have one model generate "footage" of our desired game which we could then manually modify as needed, and then have another model train on that to generate the game.
Or of course, something that skips the middle man and develops the game straight from the text. But it could very well be that the first examples of game generation could be some kludge of multiple otherwise unrelated models such as this
@@benwest5293 I had this same thought. As natural language processing improves, almost every paper on this channel becomes a skill for one general purpose AI. Once we can give verbal or text instructions to our neural networks, there's no limit to AI applications to improve workflow.
@@benwest5293 i mean why make a kludge if you can use info from the other networks that would otherwise be exposed. similar to adversarial examples the generator is allowed to back propagate on the discriminator.
@@orik737 Now we just need an AI that generates hundreds of hours of gameplay from a simple animation..
The only thing that makes me kind of less worried about these advancements as a Game Dev is that if AIs can make full fledged games. There's next to nothing else in software they couldn't do that humans could. In other words, the AI taking the job of software developers would be the least of both mine and Humanity's concerns.
the next two minute paper will be an AI that can pronounce karol's name. amazing!
Impossible, we need Saturn sized computers to do that
but just two papers down the line, we'll only need computers the size of Uranus to do this! what a time to be alive!
I had absolutely NO idea what his name was. After turning on captions and seeing it written, I went from having absolutely no idea to just essentially having no idea - a slight improvement.
@@moatddtutorials was this covered in a previous 2 minute paper?
I wonder, when they join this AI with the others already presented here, like that of alfhazero, the one that can abstract the style and even change it, it would give her to program a pac-man with a style for example sonic and still learn playing in a superhuman way. what a time to be alive.
wow! this is truly amazing! imagine generative 3D adventure games or shooters based on AI 😲
I like how your comment is 6 hours ago
ngl, i dont want shooters made by AI tho.
Imagine an ai that has the game code and files of every game ever made and then tasked to making a new game from the pieces.
@@rakeshchowdhury202 early patreon access. you can do it too ;)
@@captaingreenhat imagine a game that combines all games.
Show it half life
We can finally have half life 3
Just imagine the games we are gonna have in the future
Jeez
Imagine the ammount of computational power required and think of your comment again. It isn't that easy alright, rendering power + computational power + training power. 8 years good luck
Show it Half Life and you get Half Life, not Half Life 3.
We need to have AI write award winning novels before we can think about HL3..
@@advance512 Exactly. Also, we gotta teach it to fly airplanes and more virtual simulators first. Then we can get to teach it language processing, embed physics, and create better generative networks
@@Archimedes.5000 Shoving a hl3 into an openai input stimuli and expecting it to generate a hl3 is ridiculously insanity nothing else. We can try teaching it from low end games first. We can already see it struggles in Pokemon. Literally it can't even figure out the game workings.
TLDR; It's not a good idea to shove a whole vr game into a openai inputs. We can probably try HL along with collaborative learning to accelerate the process, else, it's gonna take lot of time
There goes my job..
And the programmers thought they were safe from AI automation... how ironic ;)
No, because the field is vastly misunderstood. It's even misunderstood by the average programmer.
That's why more and more programmers call themselves around a framwork name or language name like "I am a PHP programmer" or "I am a Spring programmer" or even "I am an Angular programmer" etc.
The real work is to explain what the client wants in a precise way so the computer can comply.
The AI will have hard time to do that kind of "translation" because it requires strong language processing skills, psychology and metacognition.
To simplify, if two persons use exactly the same words to describe what they want, the reality can be very far apart. And that's considering that those two people have similar background.
Those who knows often refer to a sentence like:
There is a big difference between what you want, what you believe you want, what you need, what you thought you asked, what you thought I understood, what I though you asked me, what I thought you intended to ask, what I thin you need, what I understood, what can be done and what will be done.
Jakobson's model of the functions of language is really the starters :-D
And that's even before touching the keyboard!
What will happen though, is that a lot of programmer; who weren't really, will be unable to do their job and will be fired.
It will not really make it easier for everybody, but actually uncover a niche. The niche that is actually here, but hidden. For some of us, that's actually very good news.
@@programaths If you believe that makes you safe from automation it just means you aren't looking far enough into the future.
Automation happens bit by bit, not all at once.
Just because the first 'programmer' AI's that are good enough to be useful won't be able to do what you're describing doesn't mean they'll never be able to do it.
That's wishful thinking.
And that's how you get caught out by this stuff.
Because you thought it was impossible, and therefore you didn't plan for the prospect that you could become redundant as well...
@@programaths Humans do this "miraculous" translation by using their own neural networks - it's just a matter of time before AI has rich enough contextual knowledge - think business domain, psychology etc. - to actually better understand what the client wants than the over-priced consultants. Two more papers down the line ;)
@@infinummjb It's more than that. There is a part that we do not even grasp as of today and this is required to move forward. It could happen by sheer emergence though.
It's in a sense miraculous. That's why we are still clueless about how we are made. Sure, neuroscientists are a thing, but they will be swift to tel that they are clueless. Even if it means they know much more than the average Joe. (Well, there is some Duning Kruger effect there)
On a simpler scale, everyone know the states of matter right ? Well, we routinely discover new states! I can't even keep the count of them :-D
Reality is that this is really a "can not say" topic.
On top of that, an AI being able to do the job of a real programmer has to be the singularity. So, if real programmers are off the job, everyone is. At least, they are the last on the list.
Programming is one of those field that is inherently interdisciplinary. Typing is really the mundane part.
Oh man, we programmers days are numbered!
@@Amipotsophspond in this case I'm more like an assembly programmer grumbling about kids these days and their high level Python. Not many assembly coding jobs around any more; we agree this will "evolve" programming as we know it to even higher levels.
sounds more like programmers will be the last, whose days are numbered
@randomguy8196 Some Jobs ARE already taken over by AI.
Exactly my thinking. This video, more than anything else, makes me want to start learning how to make AI and NNs.
@randomguy8196 Exactly. People don't understand the danger here. But none of this will matter once someone makes a true general Ai and then fucks up and it goes rogue, so at least we won't have to be jobless for long... as we'll have bigger problems.
IT'S FINALLY HERE!!!! god I've been waiting for this paper :DDDD
I appreciate how the total depth/breadth of applications for this research isn't obvious, but it seems it will be monumental.
I just wanted to say I really enjoy everything that you do. I am an avid learner and these video always satisfy the need to know more.
Mom, can I have Super Smash Bros. Ultimate?
Mom: But son, we already have Ultimate at home
Ultimate at home: Auto-generated from UA-cam footage using this AI.
Me: *About to become a video game programmer*
Neural networks: *I'm about to ruin this man's whole career*
.
.
.
Literally
must be scary
Which language are you using?
I am using C99.
Prefer to do things by hand. Much more rewarding,
the flip transitions are freaking me out
Anyone reminded of Hitchhiker's Guide to the Galaxy, where the humans created Deep Thought to find the Answer to Life, the Universe, and Everything (to which it answered 42 after thousands of millennia), and Deep Thought, when asked "well, what is the ultimate question of Life, the Universe, and Everything to which the answer is 42?" said "I don't know, but I will help you build the computer that will give you that ultimate question"? Like we're now at the point where we're writing programs that will eventually be able to write new programs with progressively less input.
"Everything truly behaves as expected." You said that right at the moment that pellets were phasing in and out of reality.
I'm excited for so many of these "2 papers later" papers. Cant wait to see what will be possible in the future!
Three papers later, we realized that we are simulated by an AI
What a time to be simulated!
So, hypothetically, I could create an animation of a game that does not exist (but that I would like to make), have the animation play alongside dummy user key input, and then the AI would be able to make the game?
That's amazing.
Does that mean i could create fictional gameplay of a game that i want to make, and the AI will code it for me? If so, could i then replace the blurry sprites with actual artwork?.
This would open up so many ways for game development.
The first part, yes, but replacing sprites is gonna be very difficult. The AI might be using 1000 sprites of the same image.
Great! Then i will be using really low res pixel art to eliminate the blurriness and reduce the number of training samples required.
They say it was the exact same but I don't know if it got the A.I of the ghosts down. If it showed that it did that would be extremely impressive since the A.I for the ghosts is all unique from each other
I said it before and will say it again. Favorite channel. This is the only piece of media I look forward to viewing each week.
I am loving your work RN
Can't wait until Vinny gets his hands on this.
And in a few, even more years, we will feed real-world data for it to build the matrix-style game.
"A new world, and new flesh, awaits you. We demand it."
Your content has been blowing my mind
I wonder if you could teach it Conway's game of life to the point that it can indefinitely and accurately iterate through a starting position it hasn't seen before
How to create the Matrix in 2050:
1. Input 5million Petabytes of video recordings from the internet
2. Run ML program
3. Wait like a month
4. Enjoy your hyper-realistic world simulation.
@SpinazFou quantum computers are only better at some task and even then it's only an improvement in the scaling of the algorithm, your not going to be able to make everything significantly faster. this is based on theory, actually computers could be even worse
@SpinazFou lookslike somebody doesn't know what a quantum computer does
@@user-vy1kz1co9p yeah. "every atom increases the performance by one bit" that would only be like what a 6x improvement on a classical computer? and they say the only problem is making a language to run in it, like we have QASM and Q# as theoretical languages for a quantum computer, we just don't have the hardware for it to be useful. @SpinazFou the biggest problem with quantum computers right now is reducing noise, the more complex the problem the more of an impact noise has on the result, but also the larger the chip the harder providing cooling is (amoung other things)
@@gwentarinokripperinolkjdsf683 Quantum Computing is better at resolving paths through algorithms than classical computing. This maybe used to create better NNs, but I'm not sure, since NNs use heuristics which are an entirely different field of mathematics. So it could potentially be an improvement but who knows.
When the first classical computers were brought out no one expected the internet or AI.
@@sangeetanarendrasingh5416 i mean we will defiantly see improvements. probably massive ones, but it will not result in "only a matter of seconds for training". we can see with out brains that quantum computers are not necessary for learning new things in minutes to hours
this channel does an amazing job.
in your words, two more papers down the line, I can only imagine the applications that an AI like this would have - so many possibilities!
What a time to be alive indeed! A few more papers down the line. And we'll be able to make computer games, simply bases on showing what should happen.
Feed it the video of RL people driving cars and their inputs AND the compilation of all the best car chasing scene of hollywood.
I suppose that before too long a detailed description of a video game idea could be input and an AI will programme it up!
This is amazing! It's always so cool seeing what AI can do.
Something that isn't clear in this video is that there's no program written here, since the neural network generates the screen in real time with controls as inputs !
So the game program is somehow the neural network itself.
who needs reverse engineering when you can just remake the game
This simulacrum of Pac-Man *offers a glimpse into the dreamworld of competitive Pac-Man players.
Honestly, wouldn't be surprising if our minds create simulations like this to train during sleep.
*at least that's what I'd like to believe.
Yep, I'll soon be out of a job thanks to AI, what a time to be alive
What a time to be alive....
Thanks for sharing.
The pacman, as well as the rendering is so impressive. The first shows, that I can model interactive systems. Could become helpful for reverse engineering. The rendering shows that neural networks find much more efficient way to calculate a near identical output.
With your channel, we seem to have arrived in the time, which Donald out of Interstellar discribes. We have the feeling, that everyday there is something new
The thought of AI being able to help supplement our skill gaps is really exciting, we've already seen the likes of DAIN helping to fill in some missing information for frame rate bumps, I'd love to see how it handles doing interpolation for more extreme examples than just bumping up 24fps up to 60fps etc.
I sure hope to be around when you talk about AGI when it finally arrives
I'd argue that the AI that can learn to play any Atari game is already pretty general. What couldn't it learn?
It's important to note that this AI doesn't actually program code. Rather, it's predicting what the next frame in a sequence of frames will look like given a set of inputs. The AI is basically making an interactive movie.
I love what you do doctor! Thank you for these videos. Have a nice day, take care!
And I'll see you next time!
Bruh. AI looks at the world and recreates the whole universe.
Now put a camera on an aquarium and see if the AI can make a program to simulate that.
*wow my idea suggestion from the last video actually exist!*
I'm speechless. I never thought narrow AI could perform something like this.
I'm really excited for the use of neural nets in the design and optimization of microchips. Hopefully it'll help us get past the electron tunneling problem as we pursue smaller lithographic features.
I'm still waiting for optical chips.
Me and the lads pirating every new release just by using gameplay footage.
I know its possible for ai to write programs. But i dont think games, or atleast original games would be for now. I also love your videos. You are amazing.
so in the future you'll be able to describe a game with something like: "third person shooter, world war 2, player is a german soldier, story is x and y"
and the intelligence will code it, using all the data it gathered thru observations
So wait, what happens if we show it videos of irl?
@@syth-1 it will produce videos of real life things aka movies
all u need to do is just make something like Scratch but for more advanced games, make people publish on your platform, enforce strictly accurate titles and descriptions, then feed all of the games to the neural network a few years down the line
In the future, you can show the AI real life footages of wars and it will generate the most realistic shooter games from that.
Jack_The_Cat next GTA confirmed
I wish he would have gone into a bit more detail on the AI's recreation of Pacman. I noticed that some of the pellets seemed to reappear after Pacman ate them. Does this mean that it was impossible to finish a level, or did the AI understand that Pacman needed to traverse nearly all of the paths, even if it didn't make the connection between pellets and levels? Also, were the AIs of the ghosts preserved, or did their strategies become much less distinct?
Every time he says "this episode was supported by weights and biases" my brain's initial reaction is "yeah the video was about ML and ML is built on the concepts of weights and biases so that makes sense"
I never understood why neural networks couldn't learn to program, very happy for this paper
AI program watching UA-cam videos of Pacman: "Yeah, I could program that..."
Amazing work!
When this video was supported by weights and biases, I felt that.
If we can recreate a rough interpretation just by looking at it, there's no reason we can't polish it to the extent that it's a perfect replica!
Damn, maybe Károly Zsolnai-Fehér is now substituted with a GAN that selects papers, recalls the history of the field and computes the deltas compared to the previously presented paper.
So it produced an interactive video, but also retained logic for the more rare states. Interesting. It's like replaying a memory of playing Pac-Man.
It'll be interesting to see if it'll ever be able to be as reliable as explicit programming, particularly in edge cases. Or maybe we'll just see genres of games where edge case reliability is less important, but I'm guessing we'll see a hybrid with explicitly programmed fall-backs.
Also, it's very interesting to follow how scalable the programming AI can be. Thanks.
Imagine a game where the AI made realistic fluid motion and light reflections
Sometimes I forget that we're living in the future...
Probably cause we are still there hoping for police to not arbitrarily kill black people. We're getting the technology from utopias SF, but the society from dystopias...
What a time to be alive!
Maybe I'm missing something but this just looks like regular neural nets, you give it the data and answers and you get back the rules (in this case the code) vs regular programming where you give it the rules and data and get back the answer.
So, Future GPU is not Graphic Processing Unit but Game Producing Utility.
This is great! Now if you can generate or cut a lot of footage from a game you envision, that doesn't exist yet, you don't have to do any programming! Clearly at the moment it's more work than just learning programming and doing it yourself but hey, 4 papers down the line and you only give it a concept idea in plain text and some concept art and the AI makes the game for you.
The idea of making an AI that writes programm code has been on my mind for quite some time now and maybe it will help us revive dead people, in much of the same way that it can create programms that run on the same base as itself.
I wonder how accurate are ghosts' AI to the original. The Ghost AI in pacman are uniquely meticulous.
This is an incredibly big step towards AGI. If the neural network could affect it's own programming, there's potential for amazing things. I still believe something like AGI is years away, but this research is shaving decades off decades of progress.
Oh my god so I could like... animate how I'd want a game to look, and a game would pop out how I want it??? That's fucking amazing.
This is imcomprehensibly possible that AI is able define its input in order to learn and produce outputs.
Okay but that slap chop ad I just seen kinda smacks.
Dude! I love my job too! I know that feeling! What a time to be Alive!
This is beautiful alchemy
I thought the AI had 5000 hours of input not 120, but i might be incorrect.
I wanna play a game with ALL the papers AI TOGETHER, in real time. Walky model with sand and water and building and smart combatants.
This is NOT AT ALL a treat for developer.
The field is vastly misunderstood. It's even misunderstood by the average programmer.
That's why more and more programmers call themselves around a framwork name or language name like "I am a PHP programmer" or "I am a Spring programmer" or even "I am an Angular programmer" etc.
The real work is to explain what the client wants in a precise way so the computer can comply.
The AI will have hard time to do that kind of "translation" because it requires strong language processing skills, psychology and metacognition.
To simplify, if two persons use exactly the same words to describe what they want, the reality can be very far apart. And that's considering that those two people have similar background.
Those who knows often refer to a sentence like:
There is a big difference between what you want, what you believe you want, what you need, what you thought you asked, what you thought I understood, what I though you asked me, what I thought you intended to ask, what I thin you need, what I understood, what can be done and what will be done.
Jakobson's model of the functions of language is really the starters :-D
And that's even before touching the keyboard!
What will happen though, is that a lot of programmer; who weren't really, will be unable to do their job and will be fired.
It will not really make it easier for everybody, but actually uncover a niche. The niche that is actually here, but hidden. For some of us, that's actually very good news.
smartest comment in the section, i couldnt agree more with you. this isnt at all a treat for programmers, but the opposite, this will bring lots of new opportunities
You mean threat?
But given a description, an AI could code thousands of solution per seconds and read the user brain wave to optimize awe.
@@SeeAndDreamify yes, I do.
@teflontelefon Can you highlight the contradiction ?
Comment actually says that it's not bothering at all for real developers.
Maybe it's the type in "treat". I meant "threat".
Goodness, I love your job, too!
I call that brute-force programming. I figured something like that was possible. Ya know, I think its possible for AI to be able to come up with its own instruction set. That would be interesting.
wait so if you just make an animation of an game you want to make but you dont know any programming can you use this?
It would be interesting to see the actual code that was written by the AI.
Finally I can teach my AI to do my work for me, what a time to be alive
This has me wondering whether AI could be used to write accurate and efficient emulators for video game preservation purposes if given enough information. If possible, that would be extremely exciting.
I'm afraid people didn't got what kind of a revolution this is, I think that we will see some radical changes in our life thanks to this amazing work.
Plus what if some day an AI could create another AI instead of a program. This process would lead to basically an AI that can learn to do every single task.
Well explained !
theese things scare me far more than horror movies